Professional Documents
Culture Documents
www.ercim.org
Special:
Grids:
The Next
Generation
CONTENTS
G
rids are an emerging technology that promises to network which links
change the lives of individuals, organisations and over 3 000 research
society as profoundly as e-mail and the Web have done institutions across
in the past decade. It is perhaps not surprising that the first Europe.
Grids were developed for the European Organisation for Fabio Colasanti, Director-General
Nuclear Research (CERN), the particle physics laboratory that The budget for Grid European Commission Directorate-
gave us the World Wide Web. Technology research in General for "Information Society"
FP6 is more than double
By providing everyone with the high-performance computing that in FP5 – a clear
and knowledge capabilities previously available only to the demonstration of the
largest corporations and laboratories, Grids have an immense field’s strategic impor-
potential to underpin sustainable growth by improving the tance. The first 12 FP6 IST Grid research projects launched in
competitiveness of existing industries and by helping usher in summer 2004 with total EU funding of 52M will create a
new markets and services previously thought impossible. It is 'critical mass' of expertise and resources from across Europe.
expected that their impact on our quality of life will be Ambitious in their scope, they aim to deliver, inter alia, new
profound, allowing us to monitor and model more successfully Grid architectures by the end of the decade, to develop generic
everything from global climate change to the way cars behave Grid technologies and to advance significantly our capabilties
in collisions. in vital areas such as security, business models, data integra-
tion, programming models, collaboration and knowledge
Grids are both a crucial enabling technology for reaching the discovery.
‘Lisbon strategy’, set in spring 2000, of transforming the
European Union into the most competitive, knowledge-based One of the largest projects in the portfolio just launched – the
economy in the world by the year 2010, as well a fundamental Network of Excellence CoreGRID – addresses longer term
building block in the realisation of the European Research Area Grid research and will play an essential role in creating the
(ERA). foundations for next generation Grids. The European Research
Consortium for Informatics and Mathematics, ERCIM, leads
In light of this significant potential, the European Commission this significant endeavour. The action will bring together
has been financing Grid research since early 2000, when the existing, but fragmented, Grid research communities. By doing
first EU-funded projects were launched under the Fifth so it will strive for durable excellence, integration and co-oper-
Framework Programme for Research and Technological ation in Grid research across Europe, thus contributing to the
Development (FP5). The Grid research projects under FP5 creation of a European Research Area for Grids.
were focused on both technology development and application
pilots. In FP6 (2002–2006) a new technology-driven approach Considering the significant and sustained European investment
to Grid research is being pursued by the European in Grid research, Europe’s strengths in this area are well estab-
Commission, substantiated by the longer-term vision of “Next lished. The challenge is now to create more effective routes for
Generation Grids” (http://www.cordis.lu/ist/grids). Research industrial exploitation in order to translate successfully
challenges to be addressed for the realisation of this vision research results into economic benefits. For Europe to capi-
include the conceptualisation and/or development of 'network- talise better on its strengths, it is indispensable that collabora-
centric Grid operating systems' as new or enhanced fabrics for tion between research organisations, funding bodies and
future distributed systems and services, scale-free, adaptive and industry at all levels of the value chain and across national
dependable Grid architectures and middleware required to borders is reinforced.
realise the 'invisible Grid' vision. In addition, research will
address new or enhanced Grids service middleware empow- More integrated, long-term research visions and effective plans
ering individuals and organisations to create, provide and use a for their implementation need to be established, taking into
variety of services, anytime, anywhere, in a transparent and account industrial needs and commercial ambitions. This will
cost-effective way. be essential for reaping benefits from the promising area of
Grid research, paving the way towards the provision of Grid,
A complementary building block to Grid Technology research software and knowledge services as a utility.
are EU-funded activities in the area of 'Research
Infrastructures' which include the deployment and testing of Fabio Colasanti
Grids over GEANT, the world’s most powerful research
3
EUROPEAN SCENE
The 8th ERCIM Workshop 'User Interfaces for All' was held in Vienna, Austria, 28-
29 June 2004, and Prof. Chris Stary, University of Linz, was the Programme Chair
and local organiser. The workshop built on the results of the seven previous
workshops.
In the tradition of its predecessors (held who talked about 'Online Communities LNCS digital library. The workshop
in Heraklion, Crete, Greece, 30-31 for All'. adjunct proceedings are electronically
October 1995; Prague, Czech Republic, available in PDF format at
7-8 November 1996; Obernai, France, 3- The workshop attracted the strongest http://www.ui4all.gr/workshop2004/pub
4 November 1997; Stockholm, Sweden, ever interest with over 140 submissions lications/adjunct-proceedings.html.
19-21 October 1998; Dagstuhl, from all over the world, covering a wide
Germany, 28 November-1 December range of topics that include novel inter- From the year 2001 onwards, the
1999; Florence, Italy, 25-26 October action paradigms and contexts of use, ERCIM Workshop 'User Interfaces for
2000; and Paris (Chantilly), France, 24- innovative concepts of universal accessi- All' takes place every two years, in alter-
25 October 2002), the 8th ERCIM bility and sociability, new modalities and nation with the newly established
Workshop 'User Interfaces for All' aimed
to consolidate recent work and to stimu- Photo taken during the
late further discussion on the state of the workshop. Standing: Ben
art in User Interfaces for All and its Shneiderman (first keynote
increasing range of applications in the speaker), Simone Stoiber
upcoming Information Society. The (workshop assistant organiser),
special theme of this year’s workshop Constantine Stephanidis
was 'User-Centred Interaction (Working Group chair) and
Paradigms for Universal Access in the Christian Stary (workshop
Information Society'. The requirement organiser). Sitting: Noelle
for User-Centred Universal Access Carbonell (Workshop PC
stems from the growing impact of the member) and Jennifer Preece
fusion of the emerging technologies and (second keynote speaker).
from the different dimensions of diver-
sity that are intrinsic to the Information
Society. These dimensions become
evident when considering the broad
range of user characteristics, the
changing nature of human activities, the
variety of contexts of use, the increasing
availability and diversification of infor-
mation, knowledge sources and e- dialogue styles, user-centered design in 'Universal Access in Human-Computer
services, the proliferation of technolog- mobile application scenarios, late- Interaction' (UAHCI) Conference, also
ical platforms, etc. The 8th ERCIM breaking empirical results with respect held every two years in the context HCI
Workshop 'User Interfaces for All' to assessing universally accessible appli- International and affiliated conferences.
focused on the new HCI challenges cations, and standardization efforts. In 2005, the 3rd UAHCI Conference will
arising from this evolution, and on how Contributions addressed not only tech- be held in Las Vegas, Nevada, USA, 22-
these affect the continuing efforts nological solutions, but also design 27 July (http://www.hci-international.org/).
towards Universal Access in the paradigms and empirical methods for
Links:
Information Society. evaluation, as well as policy develop- 8th ERCIM Workshop 'User Interfaces for All':
ments. http://www.ui4all.gr/workshop2004/
Keynote speakers were Prof. Ben ERCIM Working Group 'User Interfaces for All':
Shneiderman (University of Maryland, The official Workshop Proceedings http://www.ui4all.gr
Maryland, USA), who presented (http://www.ui4all.gr/workshop2004/pu
'Interface Design Strategies to Promote blications/lncs3196-toc.html) will be Please contact:
Learnability for All', and Prof. Jenny published by Springer as part of LNCS Constantine Stephanidis,
ICS-FORTH, Greece
Preece (University of Maryland (Lecture Notes in Computer Science) Tel: +30 2810 391741
Baltimore County, Maryland, USA), series, in the subline 'State-of-the-Art E-mail: cs@ics.forth.gr
Surveys', and will be embedded in the
ISTAG has been set up to advise the Fraunhofer ICT Group in Germany and themselves (eg to address legal issues
Commission on the overall strategy to be Prof Michael Wilson, CCLRC potentially arising from ICT research).
followed in carrying out the IST Rutherford Appleton Laboratory, UK.
thematic priority and related activities of ERCIM also provides all ISTAG The three topics addressed by working
research as well as on the orientations members with its publications. groups in 2003 reflected three stages in
with respect to the European Research the innovation cycle: IST Research
Area by helping to stimulate the corre- The objectives of the IST thematic Content which mainly developed the
sponding European research communi- priority are not purely to produce the Ambient Intelligence vision to guide the
ties. highest quality research in Europe which structure of future work programmes;
can be exploited by European compa- the implementation issues of Human
In the context of building an IST nies, but also to meet the broader objec- resources, research infrastructures,
European Research Area, a key role of tives of the European Union established funding mechanisms and partnership;
ISTAG is to reflect and advise on the by the treaties and Council of Ministers. the exploitation of research results.
definition and implementation of a Any advice which ISTAG offers the
coherent research policy in Europe. This CEC is only taken as a recommendation In 2004 the topics considered take a
policy should ensure the mastering of on priorities and can be overruled by different approach, considering the
technology and its applications and other factors. For example, ISTAG topics for research in terms of a comple-
should help strengthen industrial cautioned that the implementation of the mentary vision of the Grid, but also
competitiveness and address the main new funding instruments in the 6th considering Grand Challenges to struc-
societal challenges. It is through such an Framework Programme would not ture research objectives rather than using
exercise that recommendations on priori- achieve the objective of freeing research visions and roadmaps alone; to consider
ties and activities of Community-funded management without more independent a particular instrument for promoting
research in IST would be drawn. financial control, but the powers of the innovation through Experience and
commission under the treaties prevented Application Research Centres (EARC);
ISTAG members are appointed as indi- them from devolving more financial and to consider actions that can be taken
viduals rather than to explicitly represent control to consortia, and the resulting as EU-wide initiatives to promote and
the interests of a single group, since instruments were implemented as we advance European research and tech-
formal representation would result in too now have them. Similarly, ISTAG has nology beyond the existing IST
large a body to reach significant deci- not strongly supported proposals for programme instruments and work
sions. ISTAG has a Chairman and 30 European Technology Platforms (ETP) programme structure.
members who provide a reasonable as a suitable instrument to structure
coverage across the EU member and research and innovation in the ICT area For further information, past working
accession states (including Rumania and in FP7, although they are supported by group reports and drafts of those under
Bulgaria), companies (eg Nokia), industries with a different structure, such development are available on the ISTAG
research institutes (eg DFKI GmbH), as aerospace. web site.
standards bodies (eg W3C), universities
Link:
(eg Open University of Catalonia), ISTAG meets in plenary session quar- ISTAG web page:
national funding agencies (eg EPSRC in terly to establish working groups and http://www.cordis.lu/ist/istag.htm
the UK) and even politicians (eg a report on their progress. The topics for
member of the Parliament of Estonia). working groups are established by the Please contact:
Among these individuals two are commission in consultation with the Michael Wilson, CCLRC
Tel: +44 1235 446619
members of ERCIM institutes who can chairman, although as ISTAG gains E-mail: M.D.Wilson@rl.ac.uk
present the interests of ERCIM: the confidence through working together
chairman Prof. Encarnacao of the they are beginning to propose topics
Link:
http://www.w3.org/2004/09/W3C10.html
8
W3C Spanish Office to promote Multimodal Web Workshop
W3C Standards in Tour of Spain and Seminars
The W3C Spanish Office organizes its first "W3C Standards On 19-20 July, W3C organized a workshop on Multimodal
Tour" around different Universities in Spain from 3 to 26 Interaction, under the auspices of the IST Programme's MWeb
November 2004. The tour aims to make W3C work known, to project.
establish contacts with researchers at Universities in Spain and
to promote the use of W3C Technologies between students As the W3C Multimodal Interaction Activity is entering its
from technical Universities. third year, much progress has been made defining the W3C
MMI framework, and the workshop was organised as an oppor-
tunity for discussing W3C's current plans and for providing
feedback and suggestions for future work. Another goal of the
workshop was looking for feedback from multimodal user
communities that are currently less well represented in the
W3C Multimodal work (eg automotive industry/telematics,
home entertainment industry, healthcare, aerospace and manu-
‘W3C Standards Tour’ consists of a route with an environ- facturing) as well as feedback and input from the Multimodal
mental friendly bus, which will make stops of one or two days research community.
at the following Spanish cities: Gijón (Opening: 3 November),
Bilbao (4 - 5 November), Zaragoza (8 November), Barcelona 41 people attended the workshop and 19 papers (publicly avail-
(9 - 10 November), Valencia (11 - 12 November), Sevilla (15 - able) were presented, both from academia and the industry,
16 November), Madrid (18 - 19 November), Salamanca (22 contributing for a better understanding of each other's experi-
November), A Coruña (24 November) and Oviedo (Closing: 26 ence and requirements. The workshop concluded with separate
November). "break-out" sessions, where participants gathered to discuss
problems of more specific interest to them, such as multimodal
The bus has access facilities for people with disabilities, 14 interaction on mobile devices or advanced academic research.
multimedia equipments, where demos of W3C Technologies Many attendants also expressed a need for a multimodal inter-
will be available, projection equipment, videoconference and action authoring language.
Internet connectivity via satellite. At the same time, confer-
ences about W3C Technologies and meetings with researchers The organizers are confident that the results of these discus-
are organized at the host Universities. sions as well as the quality of the papers and presentations
made this workshop very successful. All papers and slides are
Additionally, the “First W3C Prize on Web Standardization” available online, but the organizers are still working on editing
will be held during the tour. This national prize is an initiative the minutes of the discussions, for publication in the next few
of the W3C Spanish Office to encourage W3C Standards use weeks.
and adoption within Spanish Universities. The prize will award
the best prototype, selected among ten finalist prototypes that, Multimodal Web Seminars in Madrid and Berlin
in an innovative way, integrates any of the W3C Technologies These two Multimodal Web seminars' objective is to inform
in the following fields: Semantic Web, Device Independence, European research and industry about the results of W3C’s
Voice and Multimodal Interaction. The winner will receive an Multimodal current work. They are organized in Berlin 13
applied research grant to develop a full project based on his/her October 2004 and Madrid 18 November 2004. They will use a
awarded prototype. combination of speakers from W3C Members and W3C team.
Attendance is open to the public.
Links:
http://www.w3c.es/gira/info/intro.html.en Links:
http://www.w3c.es/ MMI Workshop page:
http://www.w3.org/2004/02/mmi-workshop-cfp.html
MMI Workshop papers:
http://www.w3.org/2004/02/mmi-workshop/papers
Latest W3C Recommendation Multimodal Web Seminar in Berlin, 13 October 2004:
http://www.w3c.de/Events/2004/membermeeting.html
• 8 September 2004: Speech Synthesis Markup Language Multimodal Interaction Seminar in Madrid, 18 November 2004:
(SSML) Version 1.0; Mark R. Walker, Daniel C. Burnett, http://www.w3c.es/Eventos/2004/Noviembre/MMI/
Andrew Hunt
9
Four W3C Workshops this Fall Semantic Web Standardization
W3C Workshop on Semantic Web for Life Sciences, in Europe: SWAD-Europe
27-28 October 2004, Cambridge, USA
The W3C is organizing a workshop to discuss emerging and by Dan Brickley, Libby Miller and Kate Sharp
future applications of Semantic Web for Life Sciences (SW-
LS). Specifically, how can Semantic Web technologies such as The SWAD-Europe project nears completion. SWAD-Europe
the Resource Description Framework (RDF), Web Ontology (Semantic Web Advanced Development in Europe) is a five
Language (OWL) and domain-specific standards such as the partner collaborative EU project, part of the Information
Life Sciences Identifier (LSID) help to manage the inherent Society Technologies programme within the 5th Framework.
complexity of modern life sciences research, enable disease The project has considerably contributed to the evolution
understanding, and accelerate the development of new thera- of the Semantic Web.
pies for disease?
SWAD-Europe originated within W3C's Semantic Web
Link:
http://www.w3.org/2004/07/swls-cfp.html
Activity, extending the approach taken in previous SWAD
work to focus on collaborative work undertaken alongside
W3C's RDF (now Semantic Web) Interest Group. The project
W3C Workshop on Constraints was created to complement and feed into 'standards track' work
and Capabilities for Web Services, within W3C. SWAD-Europe is a collaboration between W3C
12-13 October 2004, Redwood Shores, CA, USA Europe, hosted by ERCIM, research institutions at CCLRC and
Various aspects of a Web service may require description. This the University of Bristol (ILRT), and the companies Stilo
includes its constraints (eg, "You must use HTTP International Plc and Hewlett-Packard Ltd.
Authentication when accessing this service") and its capabili-
ties (eg, "I can support GZIP compression"). Likewise, clients The Original Idea
accessing Web services have constraints and capabilities of SWAD-Europe was designed to support W3C's Semantic Web
their own to consider when accessing services. This Workshop initiative in Europe, providing targeted research, demonstra-
is being held to discuss the establishment of a framework for tions and outreach. The aim was to support the development and
the expression of such constraints and capabilities and their deployment of W3C Semantic Web specifications through
association using SOAP, WSDL, HTTP, etc. implementation, research and testing activities. This involved
finding and maintaining a balance between 'in-house' open
Link:
http://www.w3.org/2004/06/ws-cc-cfp.html
source tool development, community building, outreach and
evangelism, combined with research and analysis to support and
field-test Semantic Web standards. In practice, this meant that
W3C Mobile Web Initiative Workshop, more time was spent working with members of the Semantic
18-19 November 2004, Barcelona, Spain Web Interest Group community than on in-house development
W3C is thinking of launching a 'Mobile Web Initiative'. The work, since this approach helped establish a broader base for
goal of the Mobile Web Initiative would be to make Web Semantic Web development than could be accomplished within
access from a mobile device as simple, easy and convenient as a single European project. The Interest Group provided a forum
Web access from a desktop device. Initial ideas for achieving that connected SWAD-Europe (and other SemWeb-themed
this goal include developing 'best practices' documents, projects, from Europe and elsewhere) to the more formal
providing support infrastructures for mobile developers, orga- standardisation activities of the W3C.
nizing training programs for Web content providers and
creating validation and conformance testing services for Web- SWAD-Europe's work is best understood in the context of the
access from mobile devices. challenges facing the RDF project in the 1997-2001 period prior
to the creation of the Semantic Web Activity at W3C. In RDF's
Link:
http://www.w3.org/2004/09/mwi-workshop-cfp.html
early years, a number of difficulties needed addressing. At that
time, RDF technology was perceived by some as complex and
overambitious, with poorly formalised specifications. Software
W3C Workshop on Metadata for Content Adaptation, tools were relatively immature, with toolkit-specific APIs and
12-13 October 2004, Dublin, Ireland query languages, which meant that application code was often
Content metadata is a way to express the meaning of Web tightly bound to a specific RDF toolkit. The relationship
content and to map between vocabularies of the XHTML between RDF and XML in terms of both tools and standards
family of markup languages, their attributes and texts, and the was unclear. Despite the difficulties faced by RDF in this
underlying conceptual model that they represent. Content period, there was nevertheless a lot of enthusiasm from early
metadata promises the ability to go beyond sharing simple adopters of RDF which emerged in the Interest Group. The
tagged content to sharing some of the concepts behind the tags design of the project was an attempt to find a way to use the
in the content and better device independent Web access. structure of a European project to support both formal standard-
Link:
isation work at W3C, and the community fora (Interest Groups)
http://www.w3.org/2004/06/DI-MCA-WS/cfp.html
10
which can provide a home for users of Web technology to share The overarching aim of the project was to provide, through all
their experiences. appropriate means, a body of answers to questions that had previ-
ously gone unanswered, and to foster grassroots communities
The project's initial workplan included work on reviewing and within which such concerns are addressed. Amongst its many
developing tools to access, store, search Semantic Web data, themes, SWAD-Europe provided detailed answers to developer
with the goal of motivating and informing future work on RDF questions about RDF query and storage (analysis of scalability
query languages and APIs. Practical case studies and demos issues; query languages; APIs), and human-oriented classifica-
were proposed, especially in ‘near term’ areas such as thesaurus tion (SKOS for thesauri, bookmark sharing, semantic blogging,
systems and next-generation blogging tools. A fair number of etc.). The project's final workshop was on the theme of FOAF,
workshops were included, to foster developer discussion on crit- Social Networking and the Semantic Web, and illustrated some
ical 'seed applications', eg image annotation, calendaring, social of the strengths of the project, combining presentations from
networking and geo/mapping work. In addition to these efforts, academic, commercial and open source perspectives with active
the project tried to emphasise the importance not only of stan- collaborative work on tools and applications.
dards-based technology, but on technology integration. We
undertook work on linking XML to RDF, on schema languages, One lesson from the project is that it is both important and
SVG, query, XForms and Web Services. In each case, there was rewarding to provide an environment where members of the
the same underlying intent: to enter into dialogue with users of larger community that surrounds W3C can interact and collab-
W3C technology and understand how their practical problems oratively explore the practical issues around Web technologies.
might be addressed through combing RDF approaches with The formal work of the W3C is based on small, highly focussed
other, perhaps better understood, standards-based methods. The Working Groups where individuals commit a lot of time to
goal, in effect, was to help demonstrate the basics of Semantic creating new Web standards. SWAD-Europe's primary contri-
Web as a fast-maturing technology platform. Reports from each bution was to help create a supportive background environment
of these activity areas are available from the project Web site. for such work, by allowing a much larger number of geograph-
ically-dispersed individuals to participate (through email, IRC,
Before, During and After workshops and the Web) in the Semantic Web initiative. The
Throughout SWAD-Europe, we sought to build on pre-existing project was, in the Semantic Web scene, unique in its emphasis
work, in particular work which was conducted in public fora of the practical and Web aspects of 'Semantic Web' for a Web
and was supported by freely available open source implementa- developer audience. The support that SWAD-Europe provided
tions. Previous collaborative efforts (by the project partners and to the RDF and Semantic Web Interest Groups was an impor-
members of the broader community) were supported through tant exploratory step towards a model for wider participation in
SWAD-Europe and subsequently fed into more formal activi- Web standardisation work, showing that W3C's successful
ties within W3C. The project's general approach is to utilize Working Group-led approach can be complimented by a
W3C's Interest Group mechanism to bring collaborators broader participation model which allows individual
together, test their ideas, and present the results of those collab- researchers and implementors to make real contributions to the
orations in a form that can have a good prospect of adoption deployment of existing standards and to the creation of new
within the Working Group mechanisms of W3C. ones. The challenge for the future is to work towards a Web in
which all European research efforts contribute to the communi-
What difference did SWAD-Europe make? ties which underpin the evolution of Web standards.
As always with such projects, some areas proved more
rewarding than others, and the emphasis of the project evolved Groups interested in collaborating with W3C Europe on future
in response to successes and opportunities. A significant Semantic Web-related projects are invited to contact Dan
achievement of the project has been in the area of outreach and Brickley at danbri@w3.org.
community-building. The SKOS (Knowledge Organisation
Systems and the Semantic Web) work has helped to re-engage • Dan Brickley is on the technical staff of W3C, where he
the digital library/thesaurus community. The ten workshops serves as chair of the Semantic Web Interest Group and as
held as part of the project have attracted diverse participants SWAD-Europe Project Director.
from multiple countries and specialities, and from the research, • Libby Miller is the Semantic Web Group Coordinator at
open source and business communities. Other successful work ILRT, University of Bristol.
in the project has included software development, in particular • Kate Sharp is Project Manager for SWAD-Europe at ILRT,
the leading open source C implementation of RDF in University of Bristol.
Redland/Raptor; and well-crafted and appealing demonstrators
Links:
in the areas of Semantic Blogging and Semantic Portals, SWAD-Europe reports index
showing that Semantic Web applications can be simple, prac- http://www.w3.org/2001/sw/Europe/reports/intro.html
tical and easy. The pragmatic 'walk before we run' focus of the SWAD-Europe project homepage:
project was appreciated both by Semantic Web sceptics and by http://www.w3.org/2001/sw/Europe/
enthusiasts. Project members have also made substantial SKOS: http://www.w3.org/2004/02/skos/
contributions to the editing and chairing of the RDF Core stan-
dards, and later, helped to establish 'Semantic Web phase two'
groups at W3C: the Data Access and the Semantic Web Best
Practices and Deployment Working Groups.
11
Grids:
The Next Generation
by Keith Jeffery and Péter Kacsuk
The topic of Grids was last addressed in Issue 45 of ERCIM News, in April
2001. At that time the field was relatively new. In the last few years, however,
significant efforts have been made all over Europe to establish production
Grids, in order to research every aspect and potential application of Grid
systems. The goal of the current thematic issue is to show both selected
results from this research and applications of Grid systems, and to show the
goals of some newly launched national and EU projects.
The issue contains 32 papers that can be Gb/s both within the opencluster and to
divided into the following main areas: other sites. Other important issues
• Infrastructure include virtualization and automated
• Architecture installation of grid nodes. The
• Middleware Fraunhofer Resource Grid has been
• Programming running since 2002 in Germany and
• Applications combines the computing and storage
• New projects infrastructure of the six institutes of the
Fraunhofer Grid Alliance. This infras-
For rigorous research in this field, it is tructure has been successfully applied to
necessary for Grid infrastructure to be solving engineering problems such as
developed. Unfortunately, few func- casting process optimization,
tional Grid infrastructures have so far microstructure simulation and drug
been created in Europe. It is therefore design. The Hungarian ClusterGrid is a
encouraging to find four papers production grid that has been operating
reporting on working Grid infrastruc- since mid-2002, and currently involves
tures, and three others detailing plans for 1100 nodes and provides 500 Gflops of
the creation of infrastructures. The supercomputing power for Hungarian
largest production Grid infrastructure in academic users. Its novelty is that the
Europe is the LCG Grid developed by component clusters serve educational
CERN and other physics institutes in purposes during the day and are
collaboration with the EDG (European connected to the ClusterGrid during
DataGrid) project. This was successfully nights and weekends. The result is a
completed in March 2004. Grid solution with the extremely inex-
pensive annual operation cost of 40,000
The first paper describes in detail the UK Euro.
Computing Grid and its relationship with
LCG. LCG is the largest functioning grid Icatis has developed technology similar
in the world, with over 5000 CPUs and to the Hungarian ClusterGrid: during
almost 4000 TB of storage at more than cluster usage peaks, a number of idle
70 sites around the world. The second user machines can be aggregated to the
paper reports on the CERN openlab cluster. The system is designed to be
project, a collaboration between CERN transparent to the PC owners, and the
and industrial partners to further develop machines are only used at times when
the LCG Grid infrastructure using the they are likely to be idle (nights, week-
latest industrial technologies. While the ends, vacations etc).
LCG’s middleware had initially been
targeted solely for the x86 platform, the Two special-purpose Grid infrastruc-
current goal is to port the middleware to tures are under preparation and are
Itanium. Another major focus of the reported in this issue. The first is the
project is the ‘10 Gigabit challenge’, ie European Learning Grid infrastructure,
how to achieve data transfer rates of 10 which integrates service-oriented Grid
12 Grids - The Next Generation - 30 The GRASP Project: Innovative Grid Applications:
Introduction to the Special Theme Middleware for Next Generation 47 Grid-enabled Weka: A Toolkit for
by Keith Jeffery and Péter Kacsuk Grid-Based Application Service Machine Learning on the Grid
Providers by Rinat Khoussainov, Xin Zuo and
16 Net Generation Grids: The Work of Nicholas Kushmerick
by Theo Dimitrakos
the European Commission Expert
Group 32 Akogrimo — The Grid goes Mobile 48 The eMinerals Minigrid: An
by Keith Jeffery by Stefan Wesner, Theo Dimitrakos Infrastructure to Support Molecular
and Keith Jeffrey Simulation Scientists
Grid Infrastructure:
by Martin Dove
17 GridPP: A UK Computing Grid for Grid Middleware:
Particle Physics 33 A Component-Based Software 50 Virtual Vascular Surgery on the Grid
by Sarah Pearce Infrastructure for Grid Computing by Peter Sloot and Alfons Hoekstra
by Christian Pérez 51 Photorealistic Visualisation with
18 The CERN openlab for DataGrid
Applications 34 Data Access and Integration Grid-Based Technology
by Andreas Unterkircher, Sverre Jarp by Malcolm Atkinson by Mika Pennanen and Markus
and Andreas Hirstius Ylikerälä
36 Building a Multi-Domain Grid
20 Fraunhofer Grid Alliance – From by Freek Dijkstra and David Groep 53 Solving Optimization Problems with
Grid-Based e-Science to Industrial Grid-Enabled Technologies
37 Resource Management in an Optical by Enrique Alba and Antonio J. Nebro
Applications
Grid Testbed
by Franz-Josef Pfreundt, Anette 45 Bioinformatics in the Semantic Grid
by Helmut Grund and Wolfgang Ziegler
Weisbecker, Jürgen Falkner, Thilo by Kai Kumpf and Martin Hofmann
Ernst, Uwe Der, Andreas Hoheisel, 39 Grid Resource Management using
Software Agents New Projects:
Klaus-Peter Eckert and Wolfgang
Ziegler by Isaac Chao, Ramon Sangüesa and 56 CoreGRID: European Research
Oscar Oardaiz Network on Foundations, Software
21 The Production-Level Hungarian
Multi-Agent Systems for Efficient Infrastructures and Applications for
ClusterGrid Initiative 40
Quality of Service Routing in Grids Large-Scale, Distributed Grid and
by Péter Stefán
by Eric Ayienga , Bernard Manderick, Peer-to-Peer Technologies
23 Effective Aggregation of Idle by Thierry Priol
Okello-Odongo and Ann Nowe
Computing Resources for Cluster
57 GridCoord – Co-ordinating Europe-
Computing Grid Programming:
wide Initiatives in Grid Research
by Bruno Richard and Philippe Augerat 42 Enabling High-Level Access to Grid
by Marco Vanneschi
Computing Services
24 The European Learning Grid
by José Miguel Alonso, Vicente 59 TrustCoM - A Trust and Contract
Infrastructure Integrated Project
Hernández, Germán Moltó Management Framework enabling
by Pierluigi Ritrovato and Matteo Gaeta
Secure Collaborations in Dynamic
43 Trusted Grid Workflows: Assigning
26 DILIGENT: A Digital Library Virtual Organisations
‘Reputation’ to Grid Service Providers
Infrastructure on Grid Enabled by Theo Dimitrakos, Michael Wilson
by Omer Rana, Luc Moreau and Julian
Technology and Santi Ristol
Padget
by Donatella Castelli
60 Grid.it – Next Generation Grid
44 A High-Level Grid Application
Grid Architecture: Platforms and their Applications
Environment to Grid-Enable Legacy
The Cost of Quality for Cycle- by Domenico Laforenza and Marco
27 Code
Harvesting Grids Vanneschi
by Péter Kacsuk, Tamás Kiss, Ariel
by Chris Kenyon Goyeneche, Thierry Delaitre, Zoltán
29 JGrid: A Jini Technology-Based Farkas and Tamás Boczkó
Service Grid 46 JOpera: Visual Composition of Grid
by Zoltán Juhász and Gergely Sipos Services
by Cesare Pautasso
technology with semantic technology for related work led to a definition of quality major obstacle to applying Grid tech-
the creation of future learning scenarios. of service known as Hard Statistical QoS nology in industry is the lack of a Grid-
The second is the DILIGENT project, (HSQ) and an accompanying system based ASP model. The GRASP project
which will result in a digital library architecture that can support commercial is an industry-driven European research
infrastructure based on Grid-enabled service contracts. The JGrid project is an project exploring the use of the Grid
technology. This will allow members of interesting and novel approach to services concept for solving this
dynamic virtual organizations to collab- building service-oriented Grid architec- problem. To achieve this, GRASP has
orate by exploiting shared knowledge tures based on the Jini technology and developed an architectural framework
and physical resources. Java platform of Sun Microsystems. The for Grid-based Application Service
JGrid project extends the look-up service Provision and a prototype realization of
Grid architecture is another exciting of Jini with a generic Grid discovery this framework. Since mobile communi-
subject and is addressed by four papers subsystem and with a compute broker. cation technology is widely used in
in this issue. One of the most traditional Interactive computing tasks are executed everyday life, its integration with Grid
types of Grid architecture is the cycle- by compute services, jobs are executed technology will radically advance the
harvesting grid. Work at IBM Research by batch services, and storage services pervasiveness of Grid computing across
has focused on the performance issues of allow users to store and retrieve their Europe. Through the integration of these
such architectures. Their performance- data from anywhere in the world. A two technologies, the Akogrimo project
SPECIAL THEME: GRIDS — THE NEXT GENERATION
intends to create a commercial operator- and it is therefore unsurprising that two end-users much easier. The Spanish
oriented architecture and platform that papers deal with this area of research. GRID-IT project introduced a new
support the deployment of Grid services Within the framework of the German middleware design that offers an object-
worldwide. VIOLA project, new Grid middleware oriented, high-level application
components will be developed to exploit programming interface, which simplifies
Grid Middleware the capabilities of the new underlying the process of remote task execution in
Grid middleware design is a central optical network infrastructure. These Grid deployment. It is intended to give
focus of Grid research. Six papers in this new components will include a support for parameter sweep applica-
issue address this extremely important MetaScheduler (allowing co-allocation tions. The user is provided with a high-
subject. The first introduces a new of compute resources, network resources level API that exposes a very natural and
component-based middleware approach with necessary QoS or other resources convenient point of entry to the Grid
designed as the next generation of Grid like visualization devices) and services.
middleware. The new concept will allow MetaMPICH communication libraries
automatic deployment of software for distributed execution of parallel The well-known workflow concept facil-
components on large-scale distributed applications using MPI for inter-process itates Grid programming, and is handled
systems and grids while making the best communication. The GridCat group at by two papers. Both of these emphasize
use of available resources. The project UPC is working on integrating Grid the potential commercial/business appli-
involves the design of a runtime system, technology and software-agent tech- cations of grids based on their workflow
called PadicoTM, to allow components nology in order to create an agent-based concept. The first paper extends the
to exchange data independently of the Grid Resource Management system capability of workflow systems in Grid
underlying networking technologies. (GRM). Grids and agent communities environments, generally focusing on the
The appropriate data access and integra- offer two approaches to open distributed capacity to enable more reliable, trust-
tion support in the middleware is an systems. Intelligent cooperation, coordi- worthy and efficient use of Grid
important challenge. The UK’s OGSA- nation and negotiation are issues handled resources. One of the biggest obstacles
DAI system provides a framework and by software agents and are required by to widespread industrial utilization of
set of components operating across grids Grid users and Grid resource providers Grid technology is the existence of large
or Web services that deliver these mech- in order to efficiently inter-operate. numbers of applications written in
anisms reliably. Using this framework, Another project working on the merging legacy code, which are inaccessible as
many projects are building an OGSA- of Grid and agent technologies addresses Grid services. The second paper in this
DAI and extending its repertoire. It QoS problems. In their proposed archi- area introduces a new approach known
already handles a range of relational tecture, self-interested agents will as GEMLCA (Grid Execution
systems, XML databases and collections interact using electronic market-based Management for Legacy Code
of files. techniques with the goal of establishing Architecture), to deploy legacy codes as
and maintaining a certain level of QoS in Grid services without modifying the
Unsolved security issues such as authen- the Grid network. original code. Moreover, such legacy
tication and authorization remain major code services can easily be applied by
obstacles to the industrial implementa- High-Level the end-user within the P-GRADE
tion of grids. The research area of Grid Grid Programming Concepts portal, providing visual workflow
accounting and economy systems has The next field covered by this issue is construction facilities. With simple drag-
also been neglected. The NIKHEF and frequently neglected. Since there is a and-drop style programming, users can
UvA teams are jointly developing a certain gap between the end-user needs build complex workflows from
combined authentication, authorization and the access level provided by Grid GEMLCA legacy code services. These
and accounting (AAA) service tech- middleware, in many cases end-users can be run on complex Grid systems
nique. Based on this, site-local autho- must learn low-level command inter- such as GT-3 grids, while the low-level
rization and creation of virtual organiza- faces if they wish to use the Grid. details of that grid remain completely
tion will become more dynamic and Moreover they must be fully aware of hidden to the user.
easier for both users and administrators. the underlying Grid middleware
A particular advantage of this approach concepts and re-learn the concepts and Another visual service composition
is that rather than starting everything access mechanisms every time a new system is Jopera, consisting of a visual
from scratch, they are building on the generation of Grid middleware is intro- composition language and a set of inte-
Virtual Organization Management duced. grated software development tools for
Service (VOMS) developed by INFN in composing Grid services. Jopera offers
Italy. This is one of the rare cases in High-level Grid programming concepts an open, flexible platform for Grid
which one project fertilizes another. and supporting tools and portals are service composition that significantly
Unfortunately not many papers report therefore extremely important for end- facilitates the work of end-users in
such result transfer between projects. users, making access to Grid services accessing the Grid.
more convenient. Four papers of the
Resource management is a classic and current issue describe concepts, tools The final and perhaps most important
central issue in Grid middleware design and portals that will make the lives of class of papers demonstrates that the
produce photorealistic
images in a faster and Yet another EU project is TrustCom, an
Simulation of a particle physics experiment. Read about cheaper way. A new integrated project that aims to develop a
the Large Hadron Collider Computing Grid in the articles grid known as Grix was framework for trust, security and
on pages 17 and 18. developed based on contract management in dynamic virtual
platform-neutral archi- organizations. The framework will
tecture. enable collaborative business processes
to be undertaken securely in self-
Grid really is able to contribute to the Complex problems that can only be managed and dynamic value-chains of
realization of e-science. They also show solved in non-polynomial time are businesses and governments. A fourth
different application areas in which Grid becoming common in many domains of paper describes a national Grid project
technology can provide significant our lives: economy, industrial environ- from Italy. Grid.it has a strong interdisci-
support for the fast and efficient solution ments, bioinformatics, etc. For such a plinary character and aims to define,
of large scientific problems. The particle wide spectrum of problems, heuristics implement and apply innovative solu-
physics application has already been come to the rescue, since exact tech- tions for networked computing-enabling
mentioned in the context of existing niques are unable to locate any kind of platforms, which are oriented towards
production Grid infrastructures, but solution. The fifth paper addresses these scalable virtual organizations.
other application areas are also covered. issues, showing current research on
solving optimization problems with Finally, we provide a forward-looking
The first paper in this class describes Grid-enabled technology like Globus, perspective with a short paper on Next
work on Grid-enabled Weka, a widely Condor, Legion and SGE. Generation Grids, which describes the
used toolkit for machine learning and Bioinformatics has already been work of the Expert Group convened by
data mining. It is a large collection of mentioned but the last paper in this the EC and also provides the URLs to the
state-of-the-art machine-learning algo- group puts it in a new context. Fully original reports of the group. Considering
rithms written in Java. Weka contains annotated data Grid services have enor- the importance of this subject we posi-
tools for classification, regression clus- mous relevance in bioinformatics, espe- tioned this paper as the first article of this
tering, association rules, visualization, cially systems biology, due to the vast Special Theme section.
and data preprocessing. In Grid-enabled number of biochemical entities and
Weka, execution of these tasks can be interactions that need to be kept track of. Naturally, one thematic issue cannot
distributed across several computers in In order to tackle the problem, an annota- cover the entire spectrum of current and
an ad hoc Grid. The second describes the tion tool for data in Grid services has planned Grid research in Europe.
eMinerals project, one of the NERC e- been developed by the Bioinformatics However, we believe that the 32 papers
science testbed projects. The scientific Department and the Department for Web briefly introduced above give a good
aim is to use Grid technology to advance Applications of Fraunhofer Institute. overview of the major Grid research
the capabilities of molecular-scale simu- directions in Europe. From these papers,
lations for the study of environmental Such a theme as this is incomplete readers will gain an overall picture of
processes. The third application area is without a look at plans for creating the European Grid research, and can posi-
extremely important and will open new next generation of grids. Four papers are tion their own strategies on dealing with
horizons for medical treatments. reporting newly launched projects that Grid systems in the future.
Medical simulations and visualizations address a variety of aspects of future
Please contact:
typically require computational power at research. The CoreGRID Network of Péter Kacsuk, STZAKI, Hungary
a level usually unavailable in a hospital. Excellence project is probably the E-mail: kacsuk@sztaki.hu
The University of Amsterdam recently largest and most influential of these in Keith Jeffery, CCLRC, UK
demonstrated Virtual Vascular Surgery Europe. It commenced activity on 1st E-mail: K.G.Jeffery@rl.ac.uk
The Next Generation GRIDs first expert current middleware did not make it easy ware. More dramatically, the report
group (NGG1) included also from to address major requirements of Next proposes the development of a new
ERCIM: Seif Haridi (SICS) Ludek Generation Grids such as self-managing, Grids operating system with the features
Matyska (CRCIM) and Jesus Labarta self-repairing, fault-toleration and scala- of the foundationware — and possibly
(SPARCIM). The report titled 'Next bility. Clearly a further tranche of also the middleware — built-in in a
Generation Grids, European Grid research topics needed to be defined. modular, component-based fashion so
Research 2005-2010' was produced in The EC reconvened an expert group that for any device (and indeed for any
the summer of 2003 and is available (NGG2) with some of the NGG1 particular role played by a device at any
under http://www.cordis.lu/ist/grids. members (including the original 3 from one time) only the components required
The key concept was ‘the invisible ERCIM) and some new ones with a brief are loaded. The objective is a common
GRID’; the GRIDs environment should to produce a new report covering – with Grids IT surface from RFID tags through
just ‘be there’ for use by applications in the increased knowledge of a further sensors, embedded and control system
science, business, healthcare, environ- year – the R&D requirements as computers, mobile phones, PDAs,
ment, learning and culture domains. perceived now. laptops, desktops, departmental servers
and supercomputers with massive data
The FP6 Call2 proposals in the GRIDs The NGG2 report (published September stores. In fact, the vision combines
area were evaluated and 4 major — 2004 available under http://www.cordis.lu/ aspects of traditional GRID computing
CoreGRID, AkoGrimo, NextGRID and ist/grids) is quite radical; it proposes a (metacomputing) with both distributed
SIMDAT — and 8 smaller projects were new architecture stack for Grids (see and P2P (peer-to-peer) architectures
successful; ERCIM is well represented Figure). This stack has applications with many additional novel features.
throughout and CoreGRID (see article in supported by middleware (like Grid The key concept is the self-managing
this issue) is coordinated by ERCIM. services but with many new features) GRIDs environment with open standard
These projects started officially in and in turn by foundationware on top of interfaces to allow multiple commercial
summer 2004, synchronized with the existing operating systems in order to offerings.
European GRID Technology Days, enhance existing operating systems to a
September 2004. common interface suitable for The vision requires the discarding of
supporting adequately Grids middle- many long-cherished computer science
However, by early 2004 it was felt that
projects emerging from Call2, while
covering many of the issues identified in
the NGG1 report, did not cover them all.
Also, increasing experience of the use of
middleware designed for metacom-
puting (connected supercomputers) in
many other aspects of IT — eg massive
data handling — had revealed some
inadequacies. In particular, despite the
welcome evolution of GRIDs middle-
ware towards service-oriented
computing — with the adoption from the
World Wide Web Consortium web
services the concept of grid services
defined by OGSA — the architecture of A New Architecture Stack for Grids.
principles. Global state can no longer be (identification, authentication and autho- controlled process plant, advisor systems
maintained and multiple local states with rization). Synchronisation and real-time for various aspects of life and business,
an interconnection and synchronization operation is a real issue: for some appli- environmental monitoring and control
protocol is required – the so-called ‘soft cations it is essential (eg control systems, enhanced lifelong e-learning,
state’ or ‘relative state’. ACID (atom- systems) and so special protocols and intelligent financial management,
icity, consistency, isolation, durability) guaranteed services will be required. In advanced scientific systems with
properties of database transactions upon an environment with millions of modeling and activation of detectors,
which almost all business depends today connected nodes, intermittent and disaster management and many more –
are no longer tenable; it is impossible to mobile communications and semantic all of which are at best only part-realised
maintain, eg a lock for update across rich messages (with linguistic and without GRIDs. More importantly - and
multiple self-managing nodes and so cultural heterogeneity) carried within a subject to trust, security and privacy -
new transaction models, and new trans- formally-defined syntax there are plenty these systems will all interoperate as
action compensation and recovery of R&D challenges for information required, and with a common user inter-
models, are required. The ‘classical’ systems engineering. face structure, to weave the fabric of the
notions of security are also no longer knowledge society.
relevant; cooperating systems with This vision will permit the realization of
Please contact:
different security policies and protocols applications eg intelligent clothes, the Keith G Jeffery, CCLRC, UK
will have to negotiate trust arrangements intelligent home, the intelligently- E-mail: k.g.Jeffery@rl.ac.uk
in order to provide end-to-end security assisted hospital, the remotely- Tel: +44 1235 44 6103
GridPP:
A UK Computing Grid for Particle Physics
by Sarah Pearce
UK particle physicists have built a functioning prototype Grid, to analyse the data
deluge from CERN’s next particle accelerator. Over the next three years, this will
be scaled up and integrated further with other Grids worldwide, to produce the
world’s first persistent, international, production Grid.
In 2007, CERN will introduce its Large the LHC Computing Grid is the largest jobs around the grid and deal with
Hadron Collider (LHC) — the world’s functioning Grid in the world, with over issues such as security
largest particle accelerator. LHC will 5,000 CPUs and almost 4,000 TB of • deploying computing infrastructure at
allow scientists to penetrate further into storage at more than 70 sites around the sites across the UK, to build a proto-
the structure of matter and recreate the world. With up to 5,000 jobs being run on type Grid.
conditions prevailing in the early the LHC Computing Grid (LCG) simul-
universe, just after the Big Bang. But the taneously, it is becoming a true produc- The UK GridPP testbed currently
four experiments at the LHC will tion Grid. provides over 1,000 CPUs and 1,000 TB
produce more data than any previous of storage to LCG, from 12 sites in the
coordinated human endeavour — 10 A Particle Physics Grid for the UK UK. It is linked to other prototype Grids
Petabytes each year, equivalent to a GridPP is the UK’s contribution to worldwide, and has been tested by
stack of CDs twice the height of Mount analysing this data deluge. It is a collabo- analysing data from US particle physics
Everest. Careful analysis of all of this ration of around 100 researchers in 19 experiments in which the UK is involved.
complex data will be required to look for UK University particle physics groups, Several other smaller experiments have
some of the most elusive constituents of CCLRC and CERN. The six-year, £33m also started to use the prototype Grid, and
current physics, such as the Higgs project, funded by the UK Particle particle physicists are using it to run ‘data
particle and supersymmetry. Physics and Astronomy Research challenges’, that simulate the data anal-
Council (PPARC), began in 2001 and ysis needed when LHC is up and running.
To deal with this data, LHC will use has been working in three main areas: In this way, UK particle physics has
distributed computing. More than • developing applications that will allow progressed ‘from Web to Grid’.
100,000 PCs, spread at one hundred insti- particle physicists to submit their data
tutions across the world, will allow scien- to the Grid for analysis A Tiered Structure
tists from different countries to access the • writing middleware, which will This GridPP testbed is being developed
data, analyse it and work together in manage the distribution of computing on a hierarchical model, reflecting the
international collaborations. Even today, overall structure of the wider LCG
testbed. CERN provides the 'Tier-0' CPUs, which had run more than 60,000
centre, where the LHC data will be jobs.
produced. GridPP has contributed £5m
to the CERN for this, which has been GridPP is now involved in the follow-on
used to support staff and buy hardware. EGEE project (Enabling Grids for E-
The UK’s 'Tier-1' centre at Rutherford science in Europe), which aims to
Appleton Laboratory focuses on data support the European Research Area by
storage and access. In addition there are bringing together Grids from different
four smaller, regional, 'Tier-2s' in the countries and different disciplines.
UK, with a focus on provision of
computing power for generating simu- In the 1000 million short lived particles of Working beyond Particle Physics
lated Monte Carlo data and for analysis matter and antimatter are studied each Within the UK, GridPP2 is also collabo-
of data by individual physicists. In addi- year in the LHCb particle physics rating with other parts of the UK’s e-
tion, the Grid Operations Centre (GOC), experiment. In order to design the science programme, such as the National
based at RAL, monitors the operational detector and to understand the physics, Grid Service. Many of the tools developed
status of resources deployed internatio- many millions of simulated events also by GridPP could be useful for other disci-
nally through LCG and in the UK have to be produced. plines – for example, GridPP is working
through GridPP. with clinical researchers on the potential
for using its computer security tools in the
GridPP2 – The Next Phase International Collaboration health service. In addition, GridPP
The second phase of the GridPP project As well as working with other interna- members are collaborating with industry,
began on 1 September 2004. In the lead tional particle physics experiments, sharing experience of current Grid devel-
up to 2007, this will extend the UK GridPP is playing a leading role in opment issues and solutions adopted.
particle physics grid to the equivalent of European Grid projects. During its first
Links:
10,000 PCs. The infrastructure in the UK three years, GridPP personnel were inte- GridPP: http://www.gridpp.ac.uk
will be continually tested, both by current gral to the EU-funded DataGrid project, The LCG project: http://lcg.web.cern.ch/LCG/
experiments and by the LHC data chal- which brought together scientists from EGEE: http://public.eu-egee.org/
lenges to ensure that the final system is Earth observation, bio-medicine and PPARC e-Science:
ready by 2007. By the end of this second particle physics to create prototype a http://www.pparc.ac.uk/Rs/Fs/Es/intro.asp
phase of GridPP, UK physicists will be European-wide Grid. By the time of its Please contact:
analysing real data from the LHC, using final review in March 2004, EU Sarah Pearce, GridPP Dissemination Officer
the UK Grid for particle physics. DataGrid had produced around a million Queen Mary, University of London
lines of code and had a testbed of 1,000 Tel: +44 20 7882 5049
E-mail: s.pearce@qmul.ac.uk
CERN, the European Organization for around the world requires much more between CERN and industry. Its goal is
Nuclear Research, is currently than a single organization can provide. to provide CERN with the latest indus-
constructing the Large Hadron Collider CERN has therefore launched the LHC trial technology in order that it might
(LHC) which will be the largest and Computing Grid (LCG), with the anticipate possible future commodity
most powerful particle accelerator ever mission of integrating tens of thousands solutions and adjust its Grid technology
built. It will commence operation in of computers at dozens of participating roadmap accordingly. The focus of
2007 and will run for up to two decades. centres worldwide into a global openlab's research is to find solutions
Four detectors placed around the 27 km computing resource. At the time of that go beyond prototyping, and thus
LHC tunnel on the outskirts of Geneva writing, the LCG provides about 5200 provide valuable input to the LCG
will produce about 15 Petabytes of data CPUs and 7.5 TB storage at 68 sites. In project. In this article we highlight some
per year. Storing such huge amounts of this context CERN established the of our research projects.
data in a distributed fashion and making ‘CERN openlab for DataGrid
it accessible to thousands of scientists Applications’, a new type of partnership
Please contact:
Sverre Jarp,
IT Department, CERN, Switzerland
Tel: +41 22 767 4944
E-mail: Sverre.Jarp@cern.ch
In the year 2001 five Fraunhofer descriptions, resource search mecha- the user opens the Grid Job Builder – a
Institutes launched a project named nisms including the mapping of a user’s graphical workflow editing tool – from
‘I-Lab’ which was funded by the tasks to suitable grid resources, and within the portal. With this tool complex
German Ministry for Science and finally, an advanced security concept grid jobs can be created, which are then
Education (BMBF). including the respective implementation. submitted for execution by the Grid Job
The basis on which all these subsystems Handler Web service. This Web service
This project developed an Internet labo- work together is a set of resource and dynamically deploys grid workflows to
ratory software on the basis of a workflow description languages referred the hardware resources that are most suit-
Fraunhofer Computing Grid. The objec- to as the ‘Grid Application Definition able for the execution of jobs at the time.
tive of the project was to provide a user- Language’ (GADL), and a component The Grid Job Handler communicates
friendly computing grid which gives model designed for easy integration of with resource brokers and schedulers and
users access to the grid resources and legacy applications into complex grid translates the above grid workflow into
provides appropriate solutions to their workflows. GLOBUS commands so it can be run on
computing problems. The Fraunhofer the underlying grid infrastructure.
Resource Grid (FhRG) emerged from The Fraunhofer Resource Grid can be
these efforts, and substantial applications accessed in a number of ways, As shown in Figure 1, each level can in
have been running on the FhRG since depending on the user’s level of experi- principle be bypassed, which may be
2002. The FhRG Web portal can be ence. Figure 1 gives a brief overview helpful for advanced users or special
reached at http://www.fhrg.fhg.de. over the FhRG/eXeGrid architecture and application scenarios.
Today, the FhRG combines the the information flow within it.
computing and storage infrastructure of In addition to the efforts within the I-Lab
the six participating institutes: ITWM As shown, the user can access all the project, other institutes in the Fraunhofer
(Kaiserslautern), IAO (Stuttgart), FIRST grid’s services through the Web portal, Society, namely Fraunhofer SCAI and
(Berlin), IGD and SIT (both in then going through a process of ‘task Fraunhofer FOKUS, have been involved
Darmstadt) and SCAI (Sankt Augustin). mapping’ that returns appropriate grid in Grid-related efforts, with SCAI
resources to solve the task defined by the covering supercomputing problems such
Within the I-Lab project, Fraunhofer user. After having chosen the resources, as the development of meta-message
developed a middleware package based
on, but not strictly dependent on the
Globus Toolkit. The result of these
efforts was to a large extent made
publicly available in the beginning of
2004 and is distributed through the
eXeGrid Open Source Project at
http://www.exegrid.net.
The Production-Level
Hungarian ClusterGrid Initiative
by Péter Stefán
The Hungarian ClusterGrid Infrastructure introducing a layered architectural universities, high schools or even public
handles the Grid problem in a slightly model. As its name suggests, the basic libraries, serve an educational purpose in
different way to contemporary grid building blocks of the grid are PC labs an office-like environment. Whenever
systems, building up from low-level that perform dual functions. During the they are fulfilling this function – typi-
hardware to the application level by day, the labs, which are located at cally during nights and weekends – they
are used for supercomputing in a Unix causes of bottlenecks in contemporary archies, ie all pieces of binary, library,
environment. solutions is the insufficient separation of input and output files are encapsulated
user and job credentials, which yields into the structure, and at the same time
The production grid is used by monitoring and authorization problems. the job is also a temporal entity, ie a set
researchers from many scientific fields Furthermore, it is not necessary in the of operating system processes on
and their industrial partners: mathe- ClusterGrid architecture to have the different hosts and the relationships
matics, biology, chemistry, and data submitter’s user credentials configured (communication, data transfer) between
mining are a few examples of scientific at all clusters or supercomputers in the them. The runtime execution structure
areas that need large amounts of compu- grid. This gives much more freedom to provides the following features: the job
tational power and make use of such jobs traversing through different is allowed to take its complete environ-
ment to the place of the execution (even
a licence file, or organization/virtual
organization certificate), the job can be
customized, and workflow definitions
can be treated as part of the job. Sub-
jobs (ie jobs within the master job) can
be defined and executed, and meta-jobs,
such as code compilation, can easily be
treated as ordinary grid jobs.
Connection of grid clusters via MPLS VPN over the Hungarian Academic In the future, NIIF/HUNGARNET plans
Network. to improve the national production grid
in both qualitative and quantitative
terms. This means improving the
facilities with true parallel and parameter resources (in fact the job becomes an number of compute nodes to two thou-
study applications. One of the most atomic unit on which different opera- sand (or more), installing storage nodes
popular case studies is simulating the tions, such as execution, transfer, store and eliminating data network bottle-
radiation process within the heater etc can be defined). necks.The introduction of new technical
elements of a nuclear reactor. solutions such as job gateways, SOAP-
The third innovative idea is the Web- interfaced Web portals and experimental
At the technical level, there are many service transaction-based, state-full, and IPv6 grid technologies is also of key
novel technological elements introduced distributed resource broker that provides importance for the forthcoming develop-
in the ClusterGrid. The first innovative interoperable gateway functionality to ment.
feature compared to traditional grid those grid systems built on classical
Link:
systems is that the labs are connected to disciplines using XML/SOAP interface. http://www.clustergrid.iif.hu
one another through private computer The broker itself contains simple imple-
networking (see Figure), using the capa- mentation of all basic grid services, such Please contact:
bilities of the high-bandwidth Hungarian as grid information systems, job execu- Péter Stefán, Office for National Information
Academic Network in order to enhance tion systems, and file transfer subsys- and Infrastructure Development
(NIIF/HUNGARNET), Hungary
security and user confidence in the tems. E-mail: stefán@niff.hu
whole system.
The fourth innovative element is the job
The second innovative element is the use definition format that allows a job to be
of dynamic user mapping during job defined in both static and in dynamic
execution. One of the most serious terms. Jobs are built up in directory hier-
Computing clusters can be found in lots Linux. Other approaches such as PCs’ hard disks: these are not accessible
of companies and institutions today. SETI@home or XtremWeb use a while the PCs are dedicated to cluster
Researchers and engineers have high centralized distribution of computing computing. The OS and system configu-
data-processing needs and use them to tasks to Internet machines, but do not ration of a computing PC are also the
distribute large jobs to a set of homoge- offer the smoothness and ease of use of a same as a PC from the cluster, hence
neous machines. The Linux operating Linux cluster. providing homogeneity and easing
system is a de facto standard for cluster administration.
management, providing easy administra- Icatis is developing the ComputeMode™
tion, good performance and a wide soft- software suite, which smoothly handles The system is designed to be very trans-
ware support basis. Moreover, the set of this specific issue and aggregates user parent to PC owners, and the machines
available libraries and tools makes Linux machines to the corporate cluster. A are only used at times when the PC is
a good choice for scientific applications. server is installed on the customer likely to be idle (nights, weekends and
premises and keeps track of user PCs during business trips or vacation). If the
Past projects have been focusing on how running Windows. During cluster usage user returns unexpectedly while his/her
to aggregate user workstations from the peaks, a number of idle user machines PC is doing a computation, the user can
enterprise network to the clusters. In this can be aggregated to the cluster. This is claim the PC back, and it restores to the
way, the company can take advantage of done through a transparent switch of the state in which the user left it within one
the availability of the processing power PC to a secondary, protected mode from minute. This includes the user session,
of user PCs. However, the real world which it proceeds into a network boot open files, on-going programs and the
shows that most corporate users are from the ComputeMode™ server, taking desktop.
running Microsoft Windows™, making advantage of the PXE protocol. This
it difficult to aggregate user machines to patented technology provides several On the administration side,
the corporate cluster, which is based on benefits, such as the full isolation of the ComputeMode™ offers a Web adminis-
tration interface to register/unregister
machines to the system, handle specific
system image parameters and the usage
schedules for the machines (this can be
done automatically), and check usage
logs. The Job Management System
(JMS) administration for the cluster
shows additional machines in the
computing pool, and priorities can be
adjusted using the standard JMS config-
uration tools.
computational power and processing computing. Successful experiments Some Icatis customers have already
occurs faster for the user. include the development of a supercom- evaluated a ComputeMode™ prototype
puter from standard hardware compo- on their own premises. It has shown
Icatis is a young company, having been nents such as those that might be found good results for peak usage absorption
created in January 2004 after several in a typical large company. An unusual on an application for seismic terrain
years of investigation and refinement of supercomputer built from 225 standard exploration, with each job running
its offer. It was successful on the PCs (733 MHz, 256 MB RAM, Ethernet several hundreds or thousands of tasks.
commercial side: a contract has already 100) entered the TOP500 contest The full product will be released from
been signed with a major oil and gas (http://www.top500.org/) in May 2001 QA in December 2004. Among other
company and in June 2004 Icatis was and was ranked 385th worldwide for features, future versions of
elected a laureate in the ‘Innovation- supercomputing. Other successful exper- ComputeMode will have wider grid
Development’ national contest and won iments such as I-Cluster and NFSp, as capabilities, such as inter-Cluster load
a prize from the French Agency for well as the Ka-Tools developed in close balancing, and multiple administration
Innovation (http://www.anvar.fr/). partnership with Mandrakesoft roles/domains. At the same time, Icatis is
(http://www.mandrakesoft.com/) for the working on a high-end visualization
Most Icatis researchers have been MandrakeClustering product (CLIC cluster.
working within the ID-IMAG project from ID-IMAG), have built some
Link:
Laboratory (http://www-id.imag.fr/), the sound technical bases for Icatis. http://www.icatis.com/
Apache project run by INRIA, CNRS,
IMAG, and UJF. ID is a French public Icatis benefits from a strong experience Please contact:
research laboratory, which for the past in cluster computing and the Linux Philippe Augerat, Icatis CEO,
twenty years has been researching System, and has strong links with the Grenoble, France
Tel: +33 6 76 27 27 92
concepts, algorithms and tools for high- high-performance computing commu- E-mail: philippe.augerat@icatis.com
performance, parallel and distributed nity.
A semantic Grid for human learning is the vision behind the European ELeGI
Integrated Project for the implementation of future learning scenarios based on
ubiquitous, collaborative, experiential-based and contextualized learning through
the design, implementation and validation of the Learning Grid.
In recent years, teaching and learning So, the question remains – how do we the construction of knowledge and skills
practices have been mainly based on the provide better access while maintaining in the learner, rather than the memorisa-
information transfer paradigm. This or improving quality of learning through tion of information. In keeping the
focuses on content, and on the teacher as the use of ICT? The aim of ELeGI learner at the centre of the learning
the key authoritative figure who (European Learning Grid Infrastructure), process, personalisation (exploiting the
provides information. Teachers’ efforts an EU-funded Integrated Project (23 existing learner’s capability and skills)
have generally been devoted to finding partners from nine EU countries), is to and individualisation (creating and
the best ways of presenting content in promote effective human learning by adapting learning paths according to
order to transmit information to learners. demonstrating and supporting a learning learner’s preferences) become relevant
Unfortunately the current generation of paradigm shift from the current idea of aspects to be supported by technologies
‘e-Learning solutions’ has adopted the information transfer to one that focuses through the creation of the right context.
rather narrow pedagogic paradigm of on the learner and on the construction of
‘information transfer’. This occurred knowledge using experiential and The Learning Grid
simply because it is an easy way to use collaborative learning approaches in a This new vision has two strong implica-
the Web’s basic facilities. Failures, such contextualized, personalised and ubiqui- tions with respect to technology: first,
as massive drop-out rates, are usually tous way. teaching and learning will move toward
explained by a lack of staff awareness in the form of services, and second, the
the use of the Web, rather than critical In our vision, the learner has an active creation of virtual communities that
reflection on the limits of this approach. and central role in the learning process. provide services (eg contents, tutoring,
Learning activities are aimed at aiding searching for learners for sharing experi-
ences etc) will be encouraged. creation of a semantic grid for human The Learning Grid uses its knowledge
Technology must be selected according learning: the Learning Grid. (represented through OWL-S ontology)
to these implications. ELeGI will adopt a to cross-reference learner preferences
service-oriented model, which is deeply The Learning Grid is a semantically and the pedagogical model with the
intertwined with the use of semantic enriched grid that, bringing together the tools, resources and activities available
tagging and aligns itself with the global features of Grid and semantic tech- on the Grid.
community, helping to develop trends nology, represents our solution for the
such as OGSA (Open Grid Services creation of future learning scenarios. It is The Learning Grid makes available a
Architecture), WSRF and the Next based on the OGSA model, so it inherits learning scenario virtualized as a
Generation Grid. all the features of that architecture. 'Human Learning Service', with all its
'implicit knowledge' (pedagogical
An open distributed service model is As an example of the potential offered model, learning goals, resources and
based on a concept of service (which in by the Learning Grid, we provide a short activities involved, etc…) as a building
our context differs from a product) as a description of how it should support all block for the creation of more complex
kind of predefined combination of the phases of an IMS-Learning Design and interactive learning experiences
processes yielding some result (the goal specification. composed of different scenarios and
of the service) from distributed, hetero- contexts.
geneous, asynchronously communi- Authoring tools for the production of
cating and available resources. The next learning scenarios can rely on know- The Methodology and Approach
The project’s objectives will be accom-
plished following a proactive, integrated
and iterative approach. Being proactive
is essential due to the novelty of the tech-
nologies that will be adopted. A proac-
tive approach will also be adopted for
risk management. The Figure shows the
ELeGI activities and their relationships.
We have selected a set of demonstrators
and testbeds representing scientific,
social, economic and cultural cognate
areas that include both formal and
informal learning scenarios. The key
difference between testbeds and demon-
strators is that demonstrators already
exist in non-Grid-compliant forms,
whereas testbeds are principally new
departures, designed to test the ELeGI
approach from conception to implemen-
tation and evaluation.
Links:
ELeGI activities.
EleGI Web site:
http://www.elegi.org
LeGE-WG Web site:
generation of Grid solutions will increas- ledge-based decision-making systems to http://www.lege-wg.org
ingly adopt the service-oriented model suggest pedagogical models and/or IMS-LD specifications:
for exploiting commodity technologies. activities to suit the situation. This is http://www.imsglobal.org/learningdesign/
The goal of this model is to enable and done using knowledge of both the
facilitate the transformation of informa- context and the people involved in the Please contact:
tion into knowledge, by humans as well scenario (eg starting skills, personal Pierluigi Ritrovato
Centro di Ricerca in Matematica Pura ed
as – progressively – by software agents, profiles etc). Furthermore, experts can Applicata, Università di Salerno, Italy
providing the electronic underpinning use the collaborative features of the Grid Tel: +39 089 964289
for a global society in business, govern- to cooperate in modelling the scenario. E-mail: ritrovato@crmpa.unisa.it
ment, research, science, education and In this way, the Learning Grid supports
entertainment (semantic aspects). We the analysis, modelling and development
refer to these efforts as the ‘semantic phases of Learning Design documents.
grid’.
In the delivery phase, the Learning
According to this technology scenario, Design document should be understood
the research at ELeGI should address the and its content executed.
Building on the results of its past activity in both the Grid and Digital Library
domains, ERCIM has provided the framework for the setting up of an innovative
IST 6FP funded three-year Integrated Project, DILIGENT. This project, which is
coordinated scientifically by ISTI-CNR, involves 14 European partners and a
number of international observers. These research institutions and companies
will work together on the development of a digital library infrastructure, based
on Grid-enabled technology, that will allow members of dynamic virtual
organizations to collaborate by exploiting shared knowledge and physical
resources.
Research work today is often a collabo- organization will be enabled to dynami- project, which, in the next two years, will
rative effort carried out by groups of cally create and modify its own DL by deliver a Grid production infrastructure
individuals belonging to different orga- specifying a number of requirements on shared by a very large number of
nizations remotely spread worldwide. the information space (eg publishing European organizations. DILIGENT
Motivated by a common goal and institutions, content domain, document will first enrich this infrastructure with
funding opportunities, these groups type, level of replication) and on the the necessary features for creating and
dynamically aggregate into virtual orga- services (eg service type, configuration, handling an open and 'on-the-fly' modifi-
nizations (VOs) that share their lifetime, availability, response time). A able set of services as required by the DL
resources eg knowledge, experimental reliable and secure virtual DL that satis- application framework. It will then add a
results, instruments, etc., for the duration fies the given requirements will be trans- number of specific services for:
of their collaboration, creating a rich and parently instantiated and made acces- • providing access to content sources,
powerful research environment. sible to authorized users through a such as archives, metadata repositories
portal. Many virtual DLs, serving and databases
The DILIGENT project aims at different user communities, will be • implementing typical DL functions,
supporting this new research operational active on the same shared Grid resources like search, annotation, personaliza-
mode by creating an advanced know- at the same time. The composition of a tion, document visualization
ledge infrastructure that will serve the DL will be dynamic since it will depend • supporting existing applications
needs of dynamic virtual organizations. on the currently available and registered implemented by third-parties, like
This infrastructure will allow members DL resources and on many other quality video summarization, reports genera-
of VOs to access shared knowledge, parameters such as usage workload, tion, media-specific similarity search,
services and computational resources in connectivity, etc. This development etc.
a secure, coordinated, and cost-effective model will make it possible to avoid
way. heavy investments, long delays and The DILIGENT test-bed will be demon-
radical changes in the organizations strated and validated by two comple-
DILIGENT will be built by integrating setting up these applications, thus mentary real-life application scenarios:
Grid and Digital Library (DL) technolo- fostering a broader use of DLs as means one from the environmental e-Science
gies. The merging of these two different for communication and collaboration. domain and the other from the cultural
technologies will result in an innovative heritage domain. The user community of
level of functionality providing the foun- The Grid framework will also enable the the first scenario is composed by repre-
dations for next generation collaboration provision of a number of new functions sentatives of leading organizations that
environments able to serve many whose implementation has until now operate in the environmental sector.
different research and industrial applica- been limited by their high cost in terms DILIGENT will be used experimentally
tions. of computational, storage and data as a means for supporting two of the
transfer capacity, such as multimedia typical activities of this community: the
In particular, the Grid framework will document and geographic information organization of conferences and the
provide the context for implementing the processing, 3D handling, spatial data preparation of projects and periodical
notion of virtual digital libraries (VDLs), manipulation, etc. reports on specific environmental topics
ie transient, on-demand DLs based on of concern. We expect that DILIGENT
shared computational, multimedia and From the technical point of view, will facilitate the achievement of the
multi-type resources. The DILIGENT DILGENT will exploit the results of the goals of these activities and will enhance
infrastructure will maintain a network of EGEE (Enabling Grids for E-science in the quality of their results by improving
existing resources on the Grid. A virtual Europe (http://public.eu-egee.org/) accessibility, interoperability and
usability of environmental data, models, project. The system will also be used both scientists and potential DILIGENT
tools, algorithms and instruments and by experimentally for the organization of users.
integrating the located data sources with courses. Specific VDLs that address the
specialized data handling services. knowledge needs of the students will be Information on how to become a DILI-
created by re-using material maintained GENT observer and about the DILI-
The user community of the second in the registered archive resources of the GENT activities and events can be found
scenario consists of scholars, distributed DILIGENT infrastructure. on the project website.
all over the world, who have decided to
Links:
work in a three year research project in The DILIGENT consortium expects that http://www.diligentproject.org
order to set up the basis for a new the Grid-based DL infrastructure will be http://public.eu-egee.org/
research discipline that merges together adopted and extended by several other
experiences from the medical, humanity, communities in order to serve different Please contact:
social science, and communication application areas. Some of these commu- Donatella Castelli, ISTI-CNR
Scientific Coordinator
research areas. DILIGENT will be nities have already been registered as Tel: +39 050 3152902
experimented as a means to provide project observers. Regular training E-mail: donatella.castelli@isti.cnr.it
these scholars with a cost-effective sessions and workshops will be Jessica Michel, ERCIM office
instrument for setting up DLs serving conducted to disseminate the experience Administrative Coordinator
their needs despite the limited duration and results of the DILIGENT project to Tel: +33 4 92 38 50 89
and funding resources available for their E-mail: jessica.michel@ercim.org
Business models in Grid computing antee a set of statistical parameters of Stochastic QoS is a QoS parameter or set
dealing in buying and selling resources each and every contract instance. of QoS parameters that is based on statis-
across budget boundaries (within or tical properties of some QoS metric.
between organizations) are in their very The chance of achieving a good QoS Hard Stochastic QoS (HSQ) is simply a
early stages. Cycle-harvesting (or -scav- level of the delivered resource slots just QoS parameter or set of QoS parameters
enging, or -stealing) is a significant area by taking whatever comes and passing it that is based on statistical properties of
of Grid and cluster computing with soft- on (ie, basically random sampling of the some QoS metric and is guaranteed with
ware available from several vendors. underlying harvested resources) is certainty. These metrics are guaranteed
However, creating commercial contracts shown in Figure 1. The error level for each and every contract. Statistically
based on resources made available by permitted is taken as 1% of the contract this is the difference between guaran-
cycle-harvesting is a significant chal- size. As expected, larger contracts (more teeing the properties of a sample versus
lenge for two reasons. Firstly, the char- samples per quantile) make it easier to guaranteeing the properties of the popu-
acteristics of the harvested resources are attain the fixed QoS requirement. lation from which the sample was drawn.
inherently stochastic. Secondly, in a However the probability of satisfying the In business terms this means that each
commercial environment, purchasers quality requirements with random contract can have a consistent value –
can expect the sellers of such contracts to sampling is basically zero. Our analysis not just a collection of contracts or a
optimize against the quality of service is valid when quantiles are guaranteed particular provider. A typical example in
(QoS) definitions provided. These chal- for any distribution: the analysis has no conventional commodities is the specifi-
lenges have been successfully met in dependence on the underlying distribu- cation of Random Length Lumber
conventional commodities, eg, Random tion. Thus we need a systematic way to contracts on the Chicago Mercantile
Length Lumber (RLL), traded on finan- deliver HSQ – we do this by combining Exchange; in the IT world statistical
cial exchanges (the Chicago Mercantile stochastic (harvested) resources with guarantees on network usage and
(CME, www.cme.com) in this case). The directly controlled (deterministic) capacity parameters are commonplace.
essential point for creating a commer- resources.
cially valuable QoS definition is to guar-
Figure 1: Probability of fulfilling a contract (Prob[success]) for a Figure 2: Expected proportion of deterministic resources required
given distribution of slots using random sampling from the (E[R]) to guarantee a given distribution (LogNormal with mean 3
harvested resources relative to the precision with which the and variance 1.5) to different levels of precision as defined by the
distribution is defined (number of quantiles). These results are distribution’s quantiles.
valid for any distribution.
We have defined a system architecture The motivation for this work was the size of 64, was relatively inefficient,
for delivering HSQ. This comprises five desire to see whether it was possible to leading to up to 61% of the stochastic
things: an HSQ controller; a pool of turn cycle-harvested resources into resources being wasted. Including down-
stochastic (harvested) resources; a pool something commercially valuable. Was wards substitution (ie, accepting longer
of deterministic (dedicated) resources; it possible to define a contract that was time slots as valid substitutes for shorter
monitoring of the stochastic pool of valuable given that the underlying ones) reduced deterministic resource
resources; and control of the stochastic resources are inherently stochastic and requirements by roughly half. These
resource pool. The basic idea is that the that in a commercial environment we can results are robust (change <5%) for
HSQ Controller monitors the Cycle expect the provider of a QoS guarantee changes in shape and scale parameters
Harvesting System (CHS) resources, to optimize against it? We wanted to by a factor of 2 in both directions.
using the CHS Monitor, as applied to avoid all arguments based on long term
each Contract and takes one or more of reputation because we were interested in We have concluded that commercial
two possible actions: sends an artificially having valuable (tradable) individual service contracts with hard statistical
'harvested resource end' to a particular contracts for cycle harvested resources – QoS guarantees (HSQ), based on cycle-
CHS resource; or diverts contract obliga- not the question of how to create valu- harvested resources are viable both from
tions from the CHS resources to the able reputations for the providers. a conceptual point of view and quantita-
Deterministic Resources, under control tively with only small (1% – 10%)
of the Deterministic Controller, with We have shown how HSQ can be imple- requirements for deterministic (dedi-
appropriate instructions for their execu- mented for cycle-harvested resources cated) resources.
tion. using a hybrid stochastic-deterministic
Links:
system. We have also described architec- http://www.zurich.ibm.com/grideconomics
Given that resource contracts are being ture, and algorithms, to support HSQ
traded and the owner of the HSQ system contracts. We have analyzed the algo- Please contact:
sees an Ask (ie, a request for resources) rithm behavior analytically using a Chris Kenyon, IBM Research,
with a price and a deadline, what is the distribution-free approach in terms of the Zurich Research Lab., Switzerland
E-mail: chk@zurich.ibm.com
cost to support that request? If it is below expected proportion of deterministic
the price then the HSQ system owner can resources required to support a given
respond directly to the Ask. If not then HSQ level. Thus our analysis can be
the owner can post an alternative price. applied whatever the details of a partic-
The cost will consist of the quantity of ular system happen to be.
stochastic resources required and the
relative proportion of deterministic For a particular HSQ example (see
resources to meet the quality require- Figure 2) where time slot lengths were
ments together with the costs of the two log-Normally distributed we found that
types of resources. In many situations to provide hard guarantees on 8 quantiles
the major cost will be the deterministic for contract sizes of 16 to 1024 slots,
resources. from 13% to 1% additional deterministic
resources were required. Permitting
oversampling, for example for a contract
The aim of the JGrid project is to however; a programming model is programming model. It is an object-
develop a software framework for next- required that enables the creation of oriented, platform-independent language
generation Grid systems. We believe that applications that can execute in such with built-in security and error handling
future grids will be pervasive, enclosing highly dynamic environments. This features. On its own, however, it is
various software and hardware compo- model needs to support, among other unsuitable for creating dynamic systems.
nents as services, and will provide things, state-full service invocation, Jini Technology – developed by Sun
simple and seamless access to services error handling, notification mechanism Microsystems in 1999 – provides the
from anywhere at any time. Future grids based on events and security. missing features for the Java platform.
will also be used not only for high- Jini is an infrastructure and program-
performance computing tasks, but also The Java language is a good candidate ming model for creating dynamic
for carrying out business and assisting for creating grid systems and a grid distributed systems. It supports service
society and individuals in many aspects
of everyday life. These systems will Figure 1: Part of the JGrid Service
need to be reliable, to operate with Browser graphical user interface
minimal administrator supervision; self- showing discovered services.
organizing, adaptive and self-healing
behaviour is mandatory.
discovery, protocol-independent service queries. A service browser is provided to to define a workflow, which is then
access, event-based notification, find and display available services to the executed under the supervision of a
distributed resource management and user. Discovering and controlling JGrid Workflow Manager service.
security, and hence is a natural fit for services may be a tedious task. The JGrid
creating grid systems. Compute Broker service relieves users The JGrid system has been demonstrated
from system-related tasks; a job can be on several international forums and can
Currently, the main focus of the JGrid submitted directly to the broker with be downloaded from the project home-
project is on computational problems; appropriate resource requirements, and page. A national testbed is under
we have developed key grid services the broker will then discover suitable construction that will allow users to
required for the construction of compu- batch services, submit the job and super- experiment with JGrid services, develop
tational grids. JGrid users can create vise its execution. On detecting an error, and execute grid applications. Future
execute-interactive grid applications it may take corrective measures on project activities include the develop-
using Java tasks that execute on behalf of the user such as re-submitting ment of non-computational services, eg
Compute services. Tasks can be sequen- the job to another batch service; on for media streaming and processing,
tial or parallel. Parallel programs can use successful completion, the user receives database access and e-commerce and
MPI-style message passing or remote the result. The broker provides computa- business applications, as well as
method invocations for inter-process tional users with a largely simplified providing access to services from mobile
communication. Non-Java applications interaction with the grid. devices.
such as legacy Fortran and C programs
Links:
can be executed on Batch services that The JGrid project also provides develop- http://pds.irt.vein.hu/jgrid
integrate batch runtime systems (eg Sun ment support for grid application http://jgrid.jini.org
Grid Engine or Condor) into the JGrid programmers. The P-Grade graphical http://www.sun.com/jini
framework. User data can be stored in program development environment has http://www.lpds.sztaki.hu/pgrade
the grid using services that allow users to been extended in order to generate JGrid
Please contact:
store and retrieve their files and data executables from graphically designed Zoltán Juhász,
from anywhere in the world. JGrid parallel message passing programs. The University of Veszprém, Hungary
provides wide-area discovery that P-Grade system also supports the E-mail: juhasz@irt.vein.hu
extends the Jini discovery mechanism in creation of grid workflow applications
order to efficiently find grid services on on JGrid. The graphical workflow editor
the Internet using powerful search of the P-Grade portal system can be used
GRASP is an industry driven European research project exploring the use of the
Grid Services concept for providing an effective technological basis supporting
the evolution of the ASP market towards a sustainable Utility Computing model.
The intention is to improve enterprise application service provision models so
as to be able to take full advantage of the flexibility offered by Web Services and
the Grid for supporting dynamic resource allocation, life-time management of
dynamic service instances, resources integration and efficient distributed
computation.
Application service providers empowered When properly integrated into a business this framework. GRASP integrates
by GRASP middleware are no longer workflow this bundle can soften enter- concepts from application service provi-
providing a fixed solution to the customer prise borders, giving way to new, more sion and utility computing, from Service
where the business logic is hardwired in flexible ways of secure and reliable Oriented Architectures and Web
the server side code but large portions of collaboration. To achieve this, GRASP Services, and a service-oriented view of
the business logic are achieved by has developed an architectural frame- Grid computing in order to empower this
customising and combining private and work for Grid-based Application Service new generation of grid-based application
outsourced services as needed. Provision and a prototype realization of service providers.
Grasp Architecture Grid Service instances regarding QoS championed WS-* extensions (notably
Building on top of existing OGSA which are based on technical and busi- BPEL4WS, WS-Security, WS-
compliant middleware the key function- ness metrics in the vision of GRASP. SecureConversation and WS-Trust) and
ality of Location, Instantiation and • Subsystem 'Accounting' to perform the OGSI reference implementations.
Orchestration of Grid Services is charging of service usage based on Continuing its committeemen to
extended to handle other enterprise service related contracts between pioneering the convergence of Web
concepts: service users and service providers, Services and Grid Technologies for busi-
• Subsystem "Location" to enable a obviously a must for ASPs. This ness oriented Grid applications, the
dynamic search of Grid Services based GRASP subsystem works out the GRASP consortium is currently devel-
on functional, performance, techno- composite account of all services that oping a strategy for migrating its imple-
logical and business related criteria. are being executed on behalf of a mentation base-line from compliance
• Subsystem "Instantiation" to attend the client. It takes care of the collection of with OGSI standards to compliance with
properly instantiation of Grid Services raw performance/resource data WSRF standards. A preliminary assess-
performing Load Balancing and invo- (shared with SLA monitoring); ment indicates that because of the way
cation of other related subsystems merging different costing models and that GRASP architecture has been
such as Accounting or SLA moni- applying explicitly defined pricing designed (eg loose-coupling between
toring. algorithms. Grid middleware functions and Grid-
• Subsystem "Orchestration" to support • Subsystem "Security" to offer flexible aware business and generic application
the ASP in the provision of new busi- security perimeters in which GRASP components) the cost of such a migration
ness services achieved by means of the infrastructure secures complete is substantially low and we expect the
composition and coordination of Application Service (multiple next generation of GRASP middleware
existing Grid Services. It implements distributed component Grid Services). to support WSRF concepts.
the workflow that encapsulates the It handles securely life-time and addi-
Links:
business logic of the ASP application tion and expulsion of component Grid Project Website
(described as BPEL business service instances collectively http://www.eu-grasp.net
processes). executing a complete application. GGF OGSA WG:
https://forge.gridforum.org/projects/ogsa-wg
Further business functionality is Overall, GRASP has been designing and WSRF:
provided as part of the GRASP implementing business critical subsys- http://www.oasis-open.org/committees/wsrf
Framework by means of a number of tems that have not been addressed as yet OGSI.NET: http://www.cs.virginia.edu/
services classified in different subsys- in other major Grid projects including ~humphrey/GCG/ogsi.net.html
tems that address core business support EDG, Globus and Unicore.
operations including: Please contact:
• Subsystem 'SLA management' to GRASP Implementation Strategy Theo Dimitrakos, CCLRC, UK
Tel: +44 1235 446387
support a SLA management system The current prototype of the GRASP E-mail: T.Dimitrakos@rl.ac.uk
including the use of SLA templates for project middleware leverages on the use
negotiating service provision based on of an 'open standards' technology base
QoS criteria, and to monitor the fulfil- line including W3C web services stan-
ment of contract associated to each dards and Microsoft, IBM, and BEA
Mobility has become a central aspect of life for European citizens - in business,
education, and leisure. Due to rapid technological and societal changes, there
has been an astonishing growth of technologies and services for mobile users.
Large investments have been made in order to provide the necessary
infrastructures across Europe. In 2003 the number of Internet connected handsets
or mobile hosts equalled the number of fixed Internet connected. Taking into
account this evolution, Akogrimo — by leveraging the large base of mobile users
— is aiming to radically advance the pervasiveness of Grid computing across
Europe.
Akogrimo will bring together the market Scenarios dynamic virtual organizations (MDVO)
orientation and pervasiveness of mobile The realization of a challenging frame- exhibiting the ability to dynamically
communication technology in everyday work merging Grid service architectures adapt the organizational structure to
life with the promise of a dynamically with the underlying mobile network changing local situations (through
concerted use of resources and services technology requires an iterative realiza- context awareness, availability of shared
provided through Grid infrastructures. tion approach. Iterations of the frame- mobile resources). Within these
work need one or more non-trivial MDVOs, complex workflows based
By integrating the widely disjoint worlds scenarios that are able to validate the upon Grid infrastructure and Grid
of data communication, telecommunica- features and capabilities that are added to services to access data from distributed,
tion and distributed service architectures, sometimes even
the final result of the Akogrimo exercise mobile databases
will be a commercial operator-oriented will be to dynami-
architecture and platform, which cally established and
supports the deployment of Grid services processed. Of partic-
in a worldwide perspective. ular relevance is the
integration of loca-
Selected scenarios showing the poten- tion based services
tials of the innovative Akogrimo and spontaneous
approach for the quality of life of the usability.
public citizen, for business opportunities
for both small and medium enterprises as The IP(v6) capable
well as large companies including The 'Next network supporting
network providers will validate the Generation the scenarios and
Akogrimo architecture. GRIDNET'. their related require-
ments will be
The vision of Akogrimo is accordingly a constructed and
world in which: the framework. The approach within derived from the infrastructure built in
• Grid services, pervasively available, Akogrimo foresees to define, in the Stuttgart and Madrid in the context of the
are eventually meeting the ‘every- scenario specification phase, perfor- IST Moby Dick and Daidalos projects.
where at any time in any context’ mance requirements and to validate early This infrastructure integrates IP-based
paradigm versions through prototypes or ‘valida- Mobility, AAA, QoS and network secu-
• Grid services, comprising personal- tion scenarios’ testing a limited set of the rity on a pure Mobile IPv6 platform.
ized knowledge and semantics, are functionality. Following the frame-
allowing for ad-hoc, dynamic, and work’s evolution, the validation At the moment the following scenarios
possibly federated formation of progresses from system component level have been identified:
complex problem solving scenarios in to the complete demonstrator.
everyday life, in business and science - Akogrimo eLearning
to fixed, nomadic and mobile citizens The three scenarios foreseen so far for The Akogrimo eLearning scenario will
• network and service operators are able validation impose demands that firstly be embedded in the frame of the E-
to develop new business activities and reflect requirements from ‘traditional’ Learning domain. The focus in this
to provide profitable services in such Grid applications, namely access to scenario will be to build a showcase for
an integrated world based on Grid and distributed data and compute intensive new ways of learning that is made
Mobile communications concepts. services. In addition the scenarios possible by the Akogrimo infrastructure.
demonstrate the need for mobile It is planned to liaise with the Integrated
Projects ELeGI and to utilise the Mobile nursing services, hospitals, emergency between network and Grid middleware
University testbed of the Daidalos service devices and emergency stations. new kind of applications are possible
project in order to increase the impact. that attract not only researchers in
Akogrimo Disaster Handling specific domains but enable the provi-
Akogrimo eHealth and Crisis Management (DHCM) sion of services for the public citizen
The Akogrimo eHealth testbed explores Akogrimo DHCM involves incidents making a pervasive Grid possible.
Grid technology and the Mobility where various crises or disasters should
Links:
paradigm in the healthcare domain. It be handled by rescue services and other Akogrimo: http://www.mobilegrids.org
builds on previous successful results mission-critical mobile personnel, who Moby Dick: http://www.ist-mobydick.org/
from a 6-year German priority research have to collaborate within time-critical Daidalos: http://www.ist-daidalos.org
program. Target users of a Grid based and dangerous situations such as large ELeGI: http://www.elegi.org
healthcare information system are the sport events, concerts or special loca-
Please contact:
European citizens, mobile pervasive tions such as airports or railway stations. Antonio Sánchez Esguevillas, Telefónica,
demand for healthcare services (eg Project Coordinator
chronic diseases), and from the health- Conclusion Tel: +34 983 367 577
care service suppliers all types of health- Akogrimo adds a new dimension to the E-mail: ajse@tid.es
care service providing institutions and Grid. Mobility and network integration Stefan Wesner, High Performance
stationary or mobile professionals which has not been addressed so far by other Computing Center, HLRS, Stuttgart,
include healthcare advisors, pharmacies, initiatives. With a merged infrastructure Germany, Project Manager
Tel: +49711 685 4275
enabling cross-layer communication E-mail: wesner@hlrs.de
The Padico project, which commenced tructures by promoting the use of soft- with several existing open-source imple-
in 2000, is investigating different aspects ware components both for the design of mentations. We first proposed an exten-
of software-component programming to the grid middleware itself and the design sion of the distributed object model
allow the design of middleware and of applications. However, existing soft- providing distributed parallel CORBA
applications for large-scale distributed ware component models (OMG CCM, objects aiming at encapsulating parallel
systems and grids in particular. We Sun EJB, Microsoft DCOM) are unsuit- SPMD codes easily and efficiently. A
concentrated on the challenging problem able for grids since they were not parallel CORBA object is just a collec-
of designing runtime systems to support designed to handle large-scale tion of identical standard CORBA
efficient execution of software compo- distributed systems, to use various objects managed as a single entity. Data
nents within a grid. We also proposed networking technologies or to support distribution and communication between
extending an existing software compo- coarse-grain parallelism. Moreover, parallel CORBA objects is being done
nent-model to support coarse-grain automatic deployment of software transparently to the programmers. This
parallelism within a software component components on available resources was extension of the object model imple-
while maintaining scalable connections not possible. The PARIS project-team mented in PaCO/PaCO++ was later
between components. Two INRIA started several research projects to deal incorporated in GridCCM, a parallel
researchers and three PhD students, with these problems. extension to the component model
within the PARIS project-team at proposed by the OMG (CCM). It is thus
IRISA/INRIA, were involved in this One of the first projects was to extend an possible nowadays to design multi-
research. existing component model to support physics applications based on the
coarse-grain parallelism. We selected connection of parallel components, each
The motivation behind this research was CORBA from the OMG as the target of them solving a particular physics.
to handle the complexity of grid infras- model simply because it was a standard
In parallel, we worked on the design of a exploited without the use of advanced This research received financial support
runtime system called PadicoTM, to features such as parallel streams. Thanks from the ACI GRID – the French
allow components to exchange data to the PadicoTM communication frame- national program funding Grid research
independently of the underlying work, software components can activities. Researchers involved in the
networking technologies. Most of the exchange data in a very efficient way Padico project are A. Denis, S. Lacour,
runtime systems associated with existing whatever the networking technologies C. Pérez, T. Priol and A. Ribes.
component models are only able to (SAN, LAN, WAN).
Please contact:
communicate through the plain TCP/IP Christian Pérez, INRIA, France
protocol. Efficient networking technolo- One of the issues under current investi- Tel: +33 2 99 84 72 05
gies found in clusters were therefore gation is the automatic deployment of E-mail: Christian.Perez@inria.fr
inefficiently used. Similarly, high-band- software components onto grids. We
width Wide Area Networks are under- propose a network topology description
We can make decisions and discoveries with data collected within our own
business or research. But we improve decisions or increase the chance and scope
of discoveries when we combine information from multiple sources. Then
correlations and patterns in the combined data can support new hypotheses,
which can be tested and turned into useful knowledge. We see this requirement
in medicine, engineering, sciences, economics and business.
An approach is to organise a communal • Much of the data has already been they understand their domain and
data resource so that the data is main- collected under different regimes know best how to organise it for their
tained in one place, under one organisa- chosen by the groups who conducted work.
tional scheme. There are three reasons the work. • It is incompatible with human nature –
why this doesn't work well for research • To integrate the collection process people want to 'own' their own work
data: itself would lose a valuable asset – the and decide how it should be
skill and knowledge of the collectors – conducted; in health care and engi-
neering, I have seen the process of nevertheless reflect the new information. • provenance tracking, audit trails and
trying to agree on a common model go It would provide them with confidence diagnostic logging.
on indefinitely, consuming the time of and provenance information, so that they
many experts, but never converging. could use the resulting data as reliable The UK’s OGSA-DAI (Open Grid
evidence. At the same time, it would not Services Architecture Data Access and
An important trend in research is to place restrictions on the autonomy of the Integration) system provides a frame-
organise larger shared data resources, data providers, would assure them that work and set of components operating
but at the same time the number of their data was not misused and assist in across grids or web services to deliver
collections is growing rapidly. Their ensuring they gained credit for the these mechanisms reliably. Using this
diversity increases, while multi-disci- quality and content of their data. framework, many projects are building
plinary research increasingly requires on OGSA-DAI and extending its reper-
integration of data from multiple A complete realisation of a virtual data toire. It already handles a range of rela-
autonomous and heterogeneous data warehouse is beyond the current state of tional systems, XML databases and
resources managed by independent the art, instead various aspects of it are collections of files.
communities. realised. Many research teams hand-
craft a one-off solution. This does not Perhaps the most important challenges
The Data Warehouse Approach scale well, as the skilled work of that remain are:
Another approach is the data warehouse building schema and data mappings • Automatic adaptation of schema and
– much used in retail and financial deci- grows with the number and richness of data transformations in response to
sion support for example. Here, data the data sources integrated. It is vulner- changes in a data source.
from multiple data resources is copied, able to being unsustainable as those • Integrated optimisation of computa-
'cleaned' and 'integrated' under a resources evolve and rarely implements tion, data management and data move-
common schema. Typically, more data is mechanisms to ensure data access poli- ment.
added to the data warehouse from a stan- cies and data provenance tracking. • high-level presentation of data compo-
dard set of sources on a regular periodic Projects, such as myGrid and Discovery sition, operations, analyses and trans-
basis. This works well with a small set of Net craft workflows to assemble relevant formations.
stable data resources serving well- data for each analysis. The Virtual Data
defined goals. In research, there is a Technology (Pegasus, Chimera, and All three require a formal foundation,
large and growing set of resources, an DagMan) developed by the GriPhyN based on tractable mathematical models
open-ended set of goals and each source project, encourage high-level expression of sufficient scope and realism.
changes, both in content and structure, as of combined data integration and anal-
Link:
experiments and surveys progress and as ysis, enabling the underlying system to UK National e-Science Centre:
understanding and technology advances. plan and evaluate more optimally. VDT http://www.nesc.ac.uk/
Data warehousing has its place in exploits the predominant update pattern
OGSA-DAI:http://www.ogsadai.org.uk/
securing and concentrating data, but in physics data of incremental addition
encouraging and capturing multi-source against a constant schema. Projects such Please contact:
evolution is vital. It becomes part of the as BIRN and GEON at SDSC catalogue Malcolm Atkinson,
scientific communication processes and data from multiple sources, describing National e-Science Centre, Edinburgh, UK
E-mail: mpa@nesc.ac.uk
enables rapid use in other domains of their structure and data representation.
new data, new discoveries and new clas- For example, each geological survey’s
sifications. rock classification is described with an
ontology and its coordinate system is
A virtual data warehouse (I am indebted defined. Tools are provided to manage
to Mark Parsons, EPCC, Edinburgh for these descriptions and use them to
this term.) would meet this requirement. construct schema and data mappings.
It would enable researchers to combine
data from a dynamically varying set of Technical solutions to recurring
evolving data sources. It would accom- distributed tasks underlie the assembly
modate all of the diversity, handling of data in data warehouses, communal or
schema integration (the differences in shared repositories, bespoke solutions,
the way data is structured and described) catalogued registries and in advances
and data integration (the differences in towards virtual data warehouses:
the sets of values used and their repre- • data description and discovery
sentations). It would accommodate the • access, authentication, authorisation &
changes in policy of data owners, and the accounting
changes in organisation and content of • data transformation
each data source. And it would do all of • data transport
this while presenting its users with • query, selection and aggregation
understandable and stable facilities, that • data update, bulk load and archiving
It is expected that as the Grid matures, identity, and decisions can be based on LCAS is a policy-decision module for
user communities will become more attributes issued to the user by the VO. implementing site-access policies, where
dynamic, shorter-lived, and based on The only infrastructure needed to estab- the policies can be changed at run-time.
detailed policies rather than binary lish a VO is this attribute authority, and It is currently a key element in enforcing
access decisions. Policy-based coordina- at the local resource a single line of security policies in many production
tion of all resources, and minimal per- configuration suffices both to enable systems, enabling site security officers to
resource configuration is therefore access for the new community and to limit damage caused by compromised
essential. support fine-grained policy controls identities.
defined and implemented by the VO.
Currently, most Grid infrastructures use However, access control to the resource
as their core authorization system GSI, Recently, the Virtual Laboratory for e- as a whole is not sufficient. LCMAPS,
the Grid Security Infrastructure devel- Science project in the Netherlands which can parse VOMS attributes, can
oped by the Globus Alliance. GSI is adopted this model for enforcing access dynamically assign specific Unix groups
based around the Public Key controls to data sets spread across two to fine-grained attributes. These groups
Infrastructure concept, where trusted hospitals, a research institute and the are then linked to an unprivileged
third parties issue certificates to both national computer centre. account for execution. The translation
users and resources. In plain GSI, access from VOMS attributes to Unix groups is
rights are directly linked to the user's The site aspect of the system imple- performed on the fly. Wildcard VO
identity. The resource provider keeps a mented by NIKHEF consists of two group specifications give the site control
list (grid-mapfile) of names of autho- parts: the Local Centre Authorization over the number of Unix groups assigned
rized users, and associates local accounts Service (LCAS) and the Local Credential to an individual VO. Sandboxing is thus
with these users. However, it is Mapping Service (LCMAPS). achieved inside any existing Unix
necessary to make the mappings in system.
advance, as the grid-mapfile is main-
tained locally at each resource. LCMAPS supports a variety of ways to
collect VO information from traditional
Unfortunately this does not scale well, user-proxies, VOMS attributes and grid-
since the grid-mapfile must be updated mapfiles, and a wide variety of account-
continuously as users are added to the enforcement modules for both single
community or ‘virtual organization’ machines and clusters of computers.
(VO). It is also prone to privacy prob-
lems, since it forces one to publish a list Multiple Resources
of all users in a project to all partners, Computing and storage are only part of
even if the user never intends to use a the picture however, with new resource
specific resource. types such as remote instruments and
optical networks joining the grid. This
Local Credential Mapping poses new challenges for resource
The local authorization services devel- brokers when independent organizations
oped by NIKHEF mitigate these prob- are involved.
lems. The new model is based on the
Virtual Organization Management Wide-area networks in particular have
Service VOMS, an authorization- many different stakeholders. Packets
attribute service developed in the travelling from NIKHEF in Amsterdam
DataTAG project by the Italian nuclear Apart from computing services, users to, say, the Laurence Berkeley lab in San
physics institute INFN. Access controls may want to provision dedicated network Francisco will encounter five different
are no longer bound only to the user's services at the same time. providers. With valuable resources such
as optical networks - your own personal contact other services to fulfill users’ system for the European Grid
light path from A to B - access control requests. The attributes themselves may Infrastructure, deployed by two projects:
and dynamic authorization are sine qua in fact be issued by the VOMS server, Enabling Grids for E-science in Europe
non in the commercial world. thereby seamlessly integrating network, (EGEE) and the Dutch Virtual
Consequently the user running a calcula- storage and computer access. Laboratory for e-Science (VL-E).
tion in Amsterdam, with the data in San
Links:
Francisco, will need both of these The University of Amsterdam (UvA) has http://www.nikhef.nl/grid/
resources simultaneously. already demonstrated that the AAA http://www.science.uva.nl/research/air/
server is able to make authorization deci-
The AAA server, developed by the sions involving the provisioning of a Please contact:
University of Amsterdam, fills this gap dedicated network connection, even if Freek Dijkstra, Universiteit van Amsterdam,
The Netherlands
in traditional resource management. The the connection crosses multiple Tel: +31 20 5257531
AAA server is capable of taking autho- domains. E-mail: fdijkstr@science.uva.nl
rization decisions across domains, and David Groep, NIKHEF, The Netherlands
uses policy files to determine business Now, the UvA and NIKHEF have joined Tel: +31 20 592 2179
logic. It can, based on the current policy, forces to build a role-based authorization E-mail: davidg@nikhef.nl
Resource Management
in an Optical Grid Testbed
by Helmut Grund and Wolfgang Ziegler
The German Ministry for Research and Education (BMBF) launched the Vertically
Integrated Optical Testbed for Large Applications (VIOLA) project in spring 2004.
The project is managed by ‘Deutsches Forschungsnetz’ (DFN) and will run for
three years.
Emerging new technology for optical up. To allow large complex applications The expected outcomes of the project are
networks will deliver QoS and band- stressing the capabilities of the under- threefold. First, new network techniques
width far beyond the capacity and capa- lying optical network, additional grid will be deployed and tested in an optical
bilities of today’s networks. For the eval- components will be adopted, including a testbed, and along with ambitious appli-
uation of new network components and MetaScheduler (allowing co-allocation cations will provide know-how for
architecture, the integration of network of compute resources, network resources future generations of networks, espe-
techniques and applications has proven a with the necessary QoS or other cially the next generation of the German
success in former testbeds. The technical resources like visualization devices) and NREN X-WIN. Second, the enhanced
basis of the testbed is comprised of MetaMPICH communication libraries Grid middleware originating from the
optical network components connecting for distributed applications using MPI project will become useful in the
compute resources, from which a grid for interprocess communication. German e-Science Initiative D-Grid. The
based on the UNICORE system is built third important aspect of the project is
the collaboration with other projects on a
Figure 1: national, European and international
VIOLA level.
FH Rhein-Sieg,
Testbed
Sankt Augustin
Topology. The consortium of the project is led by
the DFN, ranges from research institutes
caesar, Bonn University Fraunhofer,
of Bonn Sankt Augustin and universities to the telecommunica-
tion industry, and includes Research
Centre Jülich, Fraunhofer Institute for
Scientific Computing and Algorithms
10 Gbit/s SDH
10 GE
(SCAI), Fraunhofer Institute for Media
2.5 Gbit´s (2xGE)
planned in phase 2
Communication (IMK), the Centre of
optional in the context
of collaboration of Advanced European Studies and
GSN+ and VIOLA
Research (CAESAR), RWTH Aachen
University, Bonn University, the
Location X
University of Applied Sciences Bonn-
VSITE A
+ Scheduler URIs
VSITE B
systems, ie simple agreements. The
Unicore Client MetaScheduler UNICORE system then takes responsi-
Negotiation bility for the user’s job. At the time this
AJO Initiator
article will appear, the necessary modifi-
WS-Agreements
cations of the current version of
Unicore Gateway Unicore Gateway
WS-Negotiation/
UNICORE will be almost complete, and
Agreement
evaluation of co-allocation across
multiple sites is expected to start by the
NJS NJS (Primary) NJS
end of 2004 (see Figure 2,
MetaScheduler Integration in
TSI TSI TSI
UNICORE). By the end of 2004, a stable
version of MetaMPICH will also be
available and initial tests running appli-
cations in the VIOLA middleware envi-
Local Scheduler Local Scheduler Local Scheduler
ronment will take place in early 2005.
Negotiation Negotiation Negotiation
Once MetaMPICH becomes available,
Responder Responder Responder
parallel I/O will be implemented on top
of it. The basic optical network connec-
tivity will also be operable by the end of
2004.
Figure 2: MetaScheduler Integration in UNICORE.
Future Plans
In the last third of the project, a substan-
Rhine-Sieg (FHRS), Alcatel SEL AG, resources today, whereas the resource tially revised UNICORE version based
Siemens AG and T-Systems management system for (optical) on OGSA and WSRF will be made avail-
International GmbH. switches of the testbed implements the able as a middleware layer for the
necessary protocol. The implementation VIOLA testbed. The MetaScheduling
VIOLA-Support for Set-Up of the resource management system is service as mentioned above will interact
and Operation of the Grid done by Fraunhofer IMK and Bonn directly with the new UNICORE system:
The initial testbed providing 10 Gbit/s University. The second is co-allocation the UNICORE sends a request for a WS-
connection between the sites is located in of network and compute resources for a Agreement based on the job require-
the region of Cologne-Bonn-Aachen, given task of the user. This will be done ments submitted by a user, the
with a connection to GEANT/GRANDE through implementation of WS- MetaScheduler negotiates the agree-
in Frankfurt and an extension to Bavaria Agreement and WS-Negotiation as ments through WS-Negotiation and
(see Figure 1, Testbed Topology). All proposed by the Global Grid Forum delivers the resulting WS-Agreements
nodes of the testbed clusters may inter- working group GRAAP (Grid Resource back to the UNICORE system.
connect with their Gbit interfaces Allocation Agreement Protocol).
through the underlying optical testbed Finally, another objective of VIOLA is
network. The dynamic reservation and Two requirements for local scheduling to connect to similar projects on a
allocation of bandwidth upon user systems must be satisfied: local systems European and international level; these
request makes the network technology a must handle the negotiation protocol, include the Canadian CANARIE, the
substantial building block of evolving and must be able to do advance reserva- Czech CESNET, the Polish PIONIER,
grids. Nevertheless today’s middleware tion of resources for a negotiated time- the British UKERNA and the Dutch
lacks the functionality to handle this and slot. These requirements also guarantee SURFnet. Efforts to establish these
make it available to the end user and maximum local site autonomy, since the connections and create joint projects will
his/her applications. Enhancing existing negotiation respects the local scheduling be made starting in 2005.
grid middleware ‘UNICORE’ to create a mechanisms and site policies remain
Links:
grid infrastructure based on optical fully effective. User authentication and VIOLA: http://www.viola-testbed.de
network technologies is a substantial mapping is handled by the local
WS-Agreement/WS-Negotiation:
task in VIOLA. Research Centre Jülich, UNICORE system as usual. http://forge.gridforum.org/projects/graap-wg
Fraunhofer Institute SCAI and the
UNICORE: http://www.unicore.org
University of Applied Sciences Bonn- Current State of the Project
Rhine-Sieg are responsible for this task. A Web-service-based version of the Please contact:
MetaScheduler is now ready and co-allo- Helmut Grund, Wolfgang Ziegler,
Interaction with network resources is cation has been tested locally. This first Institute for Scientific Computing and
needed on two levels. The first is version interacts with the UNICORE Algorithms SCAI, Fraunhofer ICT Group,
Germany
scheduling of network resources through client, receiving the job requirements Tel: +49 2241 14 2298/2258
dedicated (local) resource management and sending back the results of the nego- E-mail: Helmut.Grund@scai.fraunhofer.de,
systems in a similar way to compute tiations with the local scheduling Wolfgang.Ziegler@scai.fraunhofer.de
Grids and agent communities offer two approaches to open distributed systems.
Grids focus on robust infrastructure and services, and agents on the operation
of autonomous intelligent problem solvers. The interests of grids and agents
converge when efficient management of the services provided by grids comes
to the fore. Cooperation, coordination and negotiation are issues that Grid users
and Grid resource providers need to address in order to successfully interoperate.
Here we present work developed by the GridCat group in order to provide an
agent-based Grid Resource Management (GRM) system compliant with adopted
standards both for both Grid and Agent computing.
In spite of the success of Grid computing computer scientists consider Agent- Foundation (http://www.i2cat.net),
in providing solutions for a number of oriented programming to be the next addressing several aspects of Grid
large-scale science and engineering prob- paradigm in software computing. Grid computing. For example, Globus Grid
lems, the problem of GRM remains an computing pioneers, as the developers of middleware has been tested and extended,
open challenge. GRM problems include the ‘de facto’ Grid middleware Globus and a Grid testbed has been created on
the management of user requests, their Toolkit, agree with agent experts on the UPC resources. We are also involved in
matching with suitable resources through advantages that the convergence of the various research activities, including
service discovery, and the scheduling of two approaches may bring (see their deployment of applications on grids, the
tasks on the matched resources. A high recent paper: I. Foster, N. R. Jennings construction of a generic Grid Framework
level of complexity arises from the and C. Kesselman, (2004). Brain meets and Portal, as well as the current work we
various issues GRM must address: brawn: Why Grid and agents need each present here to provide an agent-based
resource heterogeneity, site autonomy other. Proc. 3rd Int. Conf. on architecture for GRM. Aside from these,
and security mechanisms, the need to Autonomous Agents and Multi-Agent GridCat is involved in a joint project with
negotiate between resource users and Systems, New York, USA) Sun Microsystems Spain, to develop a
resource providers, the handling of the nation-wide Grid Infrastructure for
addition of new policies, scalability, fault The GridCat group at the Polytechnic collaborative science projects. The busi-
tolerance and QoS, just to cite the most University of Catalonia (UPC, ness and industrial sectors are also
relevant. http://www.gridcat.org) has been actively cooperating in the design, devel-
working for the last two years on a Grid opment and deployment of grids for the
Software agents are emerging as a solu- computing domain. This will provide media sector in the Barcelona area.
tion for providing flexible, autonomous Grid services to the research community
and intelligent software components. formed around the i2CAT project and We are currently looking at the intermin-
Multiagent Systems (MAS) are gling of Grid and Agent tech-
collections of software agents nologies. We propose an envi-
that interact through ronment based on an agent
cooperation, coordination or marketplace, in which agents
negotiation in order to reach representing Grid users can
individual and common goals. detect and meet agents repre-
There are five main dimensions senting Grid services, nego-
that software computing is tiate with them for the condi-
presently trying to address and tions of the service, and
that are shared by the Agent execute the tasks on the Grid
perspective: node corresponding to the
• ubiquity service (see Figure 1). Agents
• interconnection allow flexible negotiation,
• intelligence (and learning) exchanging messages on
• delegation (autonomic behalf of the entities they
computing) represent. In our system,
• human orientation. agents can use an extra infor-
mation source for negotiations,
Agents and MAS meet these consisting of a table containing
requirements, and many Figure 1: The Overall Architecture. the possible grid configura-
Research carried out in cooperation between University of Nairobi and the Vrije
Universiteit Brussel proposes to use multi-agent systems (MAS) for provision of
QoS at the network level in a Grid environment. In the proposed architecture, self-
interested agents will interact using electronic market-based techniques with the
aim of establishing and maintaining a certain level of QoS in the Grid network.
The Computational Grid is a piece of tional applications like FTP, Telnet and of-Service (QoS). Provisioning of QoS
infrastructure formed from a variety of email but is intolerable for emerging in a network requires its definition and
computational resources interconnected real-time, multimedia and Grid applica- deployment using various mechanisms.
by wide area network (WAN) links. The tions (eg Internet telephony, video- Network QoS can be defined as ‘a set of
current Internet cannot adequately conferencing, video on-demand etc), service requirements to be met by a
support Grid services, however, because which require high throughput (band- network while transporting a flow’. The
it is only able to offer ‘best effort’ width), low latency (delay and jitter) and requirements can be qualitative or quan-
service with no resource reservation. low packet-loss rate (reliability). This titative. Qualitatively, QoS is defined
This type of service is adequate for tradi- calls for the need to provision Quality- through user perceptions and require-
ments. Quantitatively QoS is defined intelligent resource allocation while The proposed MAS-based solution will
through the specification of require- routing traffic using machine-learning be decentralized and the predominant
ments in terms of constraints on various techniques. This research will approach MAS structure will be the market; in this
quantifiable metrics such as bandwidth, this challenge through the use of multi- case the Grid environment in which there
delay, delay jitter, reliability and cost. Its agent systems (MAS). The motivation will be buyers and sellers of resources.
deployment is done through admission the use of MAS is that grids are open Link agents will be the producers of the
control, scheduling, routing, flow- systems. In these systems, control is resources in this economy. They will
control and policing. Attempts to satisfy distributed, characteristics are not known wish to maximize the income they get by
the service requirements are being made in advance, and are dynamic and hetero- selling network resources to path agents.
by IETF, IEEE etc, through Bandwidth geneous. MAS can address this problem Path agents will be the buyers of link
Overprovisioning, QoS Architectures by computing solutions locally based on resources and sellers of path resources.
and Traffic Engineering. limited information from isolated parts Negotiations between buyer and seller
of the system, and use this information in agents will be based on the market
Since ‘best-effort’ will continue to be a social way. Such locality enables economy and will be modelled using
dominant however, all QoS mechanisms agents to respond rapidly to changes in evolutionary game theory. The MAS
are layered on top of the existing Internet the network state, while their social architecture is as shown in the Figure.
rather than replacing it with a new infras- nature can potentially enable their
tructure. The reason for this dominance actions to be coordinated to achieve When a request for a service is made, the
is that the infrastructure already exists, some wider and socially desirable effect. Interface Agent will map the service
and the routing protocols and algorithms MAS have been known to solve prob- request to the QoS parameters and come
are reliable and stable for the applica-
tions it was designed for. The MAS
Architecture.
Two of the key issues in supporting these
QoS campaigns in communication
networks are QoS specification and
routing. QoS specification specifies the
requirements needed for QoS and quan-
tifies them as accurately as possible.
QoS routing on the other hand is
concerned with routing traffic such that
the QoS requirements of the carried
traffic are met. In trying to find the lems that have the property of inherent up with a quantitative value for the
‘shortest path’ through the network, the distribution (physically and geographi- service. This value will be used to derive
current Internet routing protocols cally). They also solve problems another quantitative value for the
(OSPF, RIP, BGP) have two limitations. requiring the interconnection and inter- resources. which the Path Agent will use
First, they use single objective optimiza- operation of multiple autonomous, self- to competitively negotiate for a link with
tion algorithms which consider a single interested legacy systems. the Link Agents. Upon agreement, the
arbitrary metric such as hop count or Path Agent will use this link to move to
administrative weight during the path In the proposed architecture, agents will the next node and repeat a similar
selection process. This may lead to interact in the Grid network infrastruc- process. In this way a route with the right
congestion on this path while alternate ture with the aim of establishing and QoS parameters for a service request will
paths with acceptable but non-optimal maintaining a certain level of QoS. In be formed through the network.
conditions are relatively free. Second, allocating resources to implement QoS
‘best-effort’ algorithms tend to shift in Grids, the MAS properties to be Initially, the system will be modelled
traffic from one path to a ‘better’ path exploited include the following: through simulation. A simulator repre-
whenever such a path is found. QoS- • Agency: agents act as representatives senting the current Internet will be built,
based routing should be designed to for other entities with the express and the MAS will be built on top of this
avoid these problems and in addition purpose of performing specific acts simulator as an additional layer of
should take into account the applications' that are seen to be beneficial to the control. Challenges include finding suit-
requirement and the availability of represented entity. able paradigms for agent autonomy,
network resources. However, QoS • Autonomy: agents will make multi-agent communication techniques,
routing poses several challenges that autonomous decisions based on their agent communication platform and
must be addressed to enable the support intelligence and social ability. language, and the many layers involved
of advanced services in the Internet, both • Interaction: agents will interact with that will lead to a high communication
at intra and inter-domain levels. users, other agents and the environ- and protocol overhead.
ment through coordination (this can be
Please contact:
Contemporary trends are moving cooperation or competition), commu- Bernard Manderick, Vrije Universiteit Brussel
towards the provision of QoS through nication, and negotiation. E-mail: Bernard.Manderick@vub.ac.be
The Globus Toolkit is considered the de facto standard for Grid computing.
However, its steep learning curve has discouraged its widespread adoption. In
order to overcome this disadvantage an object-oriented software has been
developed, which provides an intuitive interface that completely hides the direct
interaction with Globus.
Grid computing technology enables the programming interface which allows the the toolkit in order to achieve their
collaborative usage of computational process of remote task execution in a purpose, which generally represents
resources across different organizations. Grid deployment to be simplified. It is remote task execution. This implies that
Nowadays, the Globus Toolkit repre- intended to give support to parameter the users must concentrate on how to
sents the industrial standard for Grid sweep applications, which involve the perform the execution, rather than on
computing, and is involved in a large execution of multiple instances of a task. what to execute, which is in factwhat they
number of research projects. Even These sorts of executions are typically are interested in. Since it is the purpose of
though Globus offers all the necessary resource-starved and thus Grid the Grid to offer advantages to users, the
tools to perform Grid-based executions, computing offers significant benefits by easier it is to use this technology, the
its steep learning curve means an intense enlarging the computational capabilities sooner the benefits are achieved.
training process is required before its of a single organization with new
benefits can be fully exploited in each computational resources from abroad. Figure 2 is a diagram showing some of
particular area of interest. the most important classes of the middle-
As represented in Figure 1, the middle- ware. The user directly interacts with
In the framework of the project GRID-IT ware was developed on top of the Java instances of these intuitive classes. For
(TIC2003-01318), funded by the Commodity Grid Kit, allowing the user example, a GridTask represents an
Spanish Ministry of Science and to completely avoid direct interaction abstraction for a task that must be
Technology, a middleware has been with Globus services such as GRAM executed in the Grid, a GridResource
developed by the Networking and High (for job execution), GridFTP (for high- represents an abstraction for a computa-
performance Computing Research performance file transfer), or MDS (for tional resource in the Grid infrastructure
Group (GRyCAP) of the Valencia resource characteristics discovery). and the BasicScheduler provides
University of Technology in order to Instead, the user is provided with a high- scheduling capabilities for the allocation
provide easy and transparent access to level API that exposes a very natural and of GridTasks to GridResources.
Grid computing facilities. convenient point of entry to the Grid
services. With this middleware, the user need only
Purpose and Description describe the application and the
of the Middleware It is important to point out that, with the supporting classes involve the entire
This middleware offers an object- direct interaction with Globus, Grid users infrastructure to achieve the remote task
oriented, high-level application must combine all the services provided by execution.
Figure 2:
APPLICATIONS Simplified
class
diagram
of the
Middleware Designed middleware.
processors in the remote resource, and On the other hand, the structural analysis
refusing to be executed if these are not of buildings in civil engineering design
In
fulfilled. can require the simulation of a large
ge
Sta cut
ion number of different structural alterna-
File Exe al)
llel tion
Par
a
t (Op The middleware also provides fault- tives under diverse load conditions, in
o in
ckp
Che Out
hp w rok s
t ati o
n x
w40 00
Sta
ge tolerant dynamic scheduling capabilities order to find that which best accom-
File
for the allocation of tasks to resources. plishes with all the economic limitations,
Fault tolerance is achieved by the appli- design aspects and safety requirements.
Local Machine
cation-dependent checkpointing support,
Remote Machine
periodically retrieving the checkpoint In conclusion, the development of high-
files to the local machine, so that an level interfaces to Grid computing facili-
Figure 3: Principal steps performed to execution can be resumed in another ties greatly enhances the usability of and
achieve remote task execution. resource in the case of failure. interest in this technology, allowing
wider adoption of the Grid by non-
Applications experts.
For each task, the middleware allows the This middleware is currently being used
Link:
dependent input file set to be described by the GRyCAP in a number of compu- http://www.grycap.upv.es
and appropriately staged into the tational fields. First of all, in the area of
resource before execution. The output biomedical applications, the study of the Please contact:
files can also be specified, and they will cardiac electrical activity, especially José Miguel Alonso, Vicente Hernández,
be transferred back to the local machine under pathological conditions, requires Germán Moltó, Universidad Politécnica
de Valencia/SpaRCIM, Spain
after execution. The task may also the execution of several parametric Tel: +34 963 87 73 56
specify an a priori quality of service, simulations in order to analyse, for E-mail: {jmalonso,vhernand,gmolto}@dsic.upv.es
requesting a minimum amount of avail- example, the influence of certain anti-
able RAM or a minimum number of arrhythmic drugs.
Creating applications by combining In automated discovery techniques, at Cardiff University. The project builds
independently-managed (often services have a rich semantic description on the development of Mathematical Web
distributed) services forms a key theme and a matchmaking algorithm attempts to Services in the European Commission
in Grid computing. Such services may be match the service request with service funded MONET (http://monet.nag.co.uk/)
hosted by different institutions, and advertisements. The match is then classi- project. The GENSS project addresses the
utilise a variety of compute and data fied as an exact match, or several flavours creation of matchmaking techniques to
resources. Workflow tools play an of ‘relaxed’ match. It is important to note allow users to discover Mathematical
important role in helping compose these that often ‘automated’ discovery also services described in terms of an
services, and considerable effort has requires having a human in the loop — as OpenMath-based ontology. Mechanisms
been put into creating workflow tools it is seldom possible to specify user needs to allow such services to be delivered as
that combine ease-of-use with support in an exact way (often users are also not part of a workflow for mathematical
for the widest range of Grid middleware. fully aware of their requirements). It is problem analysis is also being undertaken
Most Grid-centric workflow tools intended that the identification of ‘suffi- in the project. Such service discovery, by
assume the existence of a static set of ciently similar’ services to a query is often necessity, needs to be problem-specific.
services, available at the time of likely to lead to a discovery of more suit- By focusing on mathematical services, we
launching the tool. To exploit fully the able services. The ‘Grid Enabled intend to address the requirements of a
dynamic aspects of Grid infrastructure, it Numerical and Symbolic Services’ much wider range of users in science and
is necessary to allow for dynamic (GENSS) project is an EPSRC-funded engineering —than perhaps by focusing
services to be deployed and discovered. project between the University of Bath on a particular application domain (such
and the Welsh eScience Centre (WeSC) as BioSciences, for instance).
Once suitable services have been discov- University of Southampton (EPSRC- and reputation are related to each other,
ered, it is often difficult for the user to funded) and especially how a reputation index’
ascertain if the service selected will • SERENITI (SERvice-ENabled may be assigned to a group of services,
perform as expected. This introduces the Interactive Trust Informatics), led by based on the reputation of each service
issues of trust and reputation in services WeSC. within the group. The key theme in the
made available over Grid infrastructure. project is the observation that in a work-
Trust issues may be considered from the The PASOA project is investigating how flow session there is often a set of ‘crit-
service provider's or the service provenance data may be associated with ical’ services which must perform
consumer's perspective. For the a workflow process (process prove- successfully. These critical services are
provider, trust has always been associ- nance) and with each service (service also only available from a restricted set
ated with security -- focusing on authen- provenance). The problems of deter- of providers (such as a particular
ticating and authorising users accessing mining the origin of a result, especially numeric solver that can only execute on
the provider's resources. For the when it involves processing through a a restricted set of platforms). Allocating
consumer, trust is considered as a number of stages, or deciding when a reputation-index to such critical
predictability issue addressing concerns results of analysis are no longer valid services, and also determining how this
such as which service provider is truly become important concerns in open Grid value changes over time is a key factor in
reliable in delivering a service, and how environments and in establishing the assessing risk for the entire workflow.
much credence should be given to that quality and reproducibility of scientific
Links:
provider to deliver the service as adver- applications. The PASOA project there- GENSS: http://genss.cs.bath.ac.uk/
tised. The concepts of trust and reputa- fore is aiming to define execution and PASOA: http://www.pasoa.org/
tion are complex and many-faceted service provenance in relation to work-
issues — and associated with themes flow enactment, to enable better use of Please contact:
such as risk, competence, beliefs/percep- this data for reasoning about the Omer F. Rana, Cardiff University, Wales, UK
Tel: +44 29 2087 5542
tions, utility/benefit and expertise. Two ‘processes’ involved in a scientific E-mail: o.f.rana@cs.cardiff.ac.uk
projects at WeSC are attempting to collaboration over the Grid.
Julian Padget, University of Bath,
address these concerns: E-mail: jap@cs.bath.ac.uk
• PASOA (Provenance Aware Service The SERENITI project on the other hand
Luc Moreau, University of Southampton, UK
Oriented Architecture), led by the is exploring how the concepts of trust E-mail: l.moreau@ecs.soton.ac.uk
There are many efforts worldwide to tions. The primary goal of the research GEMLCA is a Grid architecture devel-
provide new Grid middleware concepts teams of University of Westminster and oped by the research team at the Centre
for constructing large production grids. SZTAKI within the UK e-Science for Parallel Computing, University of
As a result, the Grid community is at OGSA Testbed Project was to construct Westminster, and aims to make legacy
thestage of producing third-generation a high-level Grid application environ- code programs available as Grid services
Grid systems represented by the OGSA ment where end-users can: with user-friendly interfaces and without
(Open Grid Services Architecture) and • deploy and use any legacy code re-engineering the original code. Using
WSRF (Web Services Resource program as a Grid service when GEMLCA, legacy code written in any
Framework) standards. On the other creating Grid applications language (Fortran, C, Java etc) can easily
hand relatively little attention has been • easily and conveniently create be deployed as an OGSA Grid service
paid to how end-users can survive in the complex Grid applications by without any user effort. GEMLCA has
rapidly changing world of Grid genera- accessing a Grid portal. also been integrated with the P-GRADE
Figure 1:
GEMLCA conceptual
architecture.
Grid portal and workflow solutions Our goal was to enable end-users to were analysed by a program also running
(developed by SZTAKI), allowing the apply legacy code as OGSA Grid at Westminster. The applications were
end-user to construct and execute services in workflows and to provide a legacy C codes transferred into Grid
complex Grid workflows, using a user- user-friendly interface through which to services using GEMLCA. The workflow
friendly Web interface, on which legacy do this. In order to achieve our goal and visualization of execution is shown
components of the workflow show up as GEMLCA has been integrated with the in Figure 2.
Grid services. P-GRADE Grid portal solution. The
workflow editor of P-GRADE portal Future work will include the support of
GEMLCA is a general solution for the enables the connection of component other service-oriented Grid architectures
deployment of existing legacy code jobs into complex workflows through an like WSRF and pure Web services. It is
applications as Grid services without easy-to-use graphical environment. The also envisaged that plug-ins for applica-
code re-engineering. The only access Workflow Manager of P-GRADE portal tion-specific visualizers to the
point for a client to the GEMLCA archi- takes care of executing such workflows P-GRADE portal will be developed.
tecture is the front-end layer composed on various Grid systems in a user-trans-
Links:
of a number of Grid services. These parent way. The integration of GEMLCA: http://www.cpc.wmin.ac.uk/
services offer interfaces in order to GEMLCA and the P-GRADE portal ogsitestbed/GEMLCA/
deploy, query, submit, check the status allows the user to construct and execute P-GRADE:
of, and get the results back from legacy workflows from legacy codes deployed portalhttp://www.lpds.sztaki.hu/pgportal/
computational jobs. As with any other as OGSA Grid services.
Grid Service, the front-end layer is Please contact:
described in Web Services Description In order to demonstrate the capabilities Péter Kacsuk, SZTAKI, Hungary
Tel: +36 1 279 6064
Language (WSDL) and can be used by of the solution, a complex workflow E-mail: kacsuk@sztaki.hu
any Grid service client to bind and make simulating and analysing urban car
Tamás Kiss, Cavendish School of Computer
use of its functionality through Simple traffic was created and executed on five Science, UK
Object Access Protocol (SOAP). different sites of the UK OGSA Testbed. Tel: + 44 20 7911 5000, ext 3542
GEMLCA clients can be easily created Simulators run in parallel at SZTAKI, E-mail: kisst@wmin.ac.uk
using either the provided universal CCLRC Daresbury Laboratory, and
GEMLCA Stubs for Java or the WSDL Westminster, Portsmouth and Reading
file. The conceptual architecture of Universities. The simulators were fed by
GEMLCA is shown in Figure 1. a road-network generator and the results
Figure 2: Complex workflow and execution graph for analysing road traffic.
The recent shift to service-based Grids enables the use of service composition
tools for rapidly building and efficiently running distributed computations. At ETH
Zurich, we have developed JOpera, a visual composition language and a set of
integrated software development tools for composing Grid services. The system
can be freely downloaded and has been successfully tested with data-intensive
bioinformatics applications as well as large-scale, Monte-Carlo network
simulations.
Early Grid middleware and tools empha- computing. For example, large computa- JOpera offers an open, flexible platform
sized low-level resource management tional jobs are typically partitioned into for Grid service composition. It main-
aspects. Today, Grid computing has smaller units that can then be executed in tains a library of re-usable components,
adopted the notion of ‘services’ as a a pipelined, parallel or sequential which can be dragged, dropped and
basic building block for large-scale fashion. JOpera includes simple primi- connected into a data and control-flow
scientific computations. The focus has tives for these operations and allows the diagram. JOpera makes very few
moved from CPU sharing to defining programmer to precisely control what assumptions on the properties of these
computational resources and function- happens in the case of failures. JOpera components, so that the user can freely
ality as composable services. Thanks to can also suggest the most optimal parti- choose to compose the most appropriate
standards such as Web services, an tioning strategy based on the available kind of services in terms of per-
increasingly large number of basic Grid resources. formance, reliability, security and
services are being published. They convenience. Such components may
encapsulate both algorithms and data represent the invocation of basic,
sources, hiding the complexities and remote Grid services, but also, for
peculiarities of the underlying plat- example, job submissions to external
forms. With services such as basic Grid Service Library resource management and scheduling
building blocks, it becomes possible ComputeDistances DownloadInputData UploadResult CalcStatistics
systems or the execution of local
to rapidly build Grid applications by applications under a variety of oper-
composing these services at a high ating systems. Additionally,
level of abstraction. Visual Composition
composite Grid services can be re-
GridService Input
systems, such as development environ- applications, thereby supporting the the system with a controller in charge of
ments or application servers. An open development of application-specific determining the optimal configuration of
API is a key feature for providing access features built by composing existing the system based on the current work-
to composite Grid services in a variety of Grid services. Additionally, we have load.
ways. In the simplest case, once Grid developed generic, visual monitoring Link:
services are composed, the resulting tools for keeping track of the progress of The latest version of JOpera can be freely
computation can also be published as a a Grid computation. In order to handle downloaded from http://www.iks.ethz.ch/jopera
Grid service. To do so, the JOpera kernel large workloads, the JOpera kernel can
Please contact:
manages the execution of the composite scale across a cluster of computers. We
Cesare Pautasso, ETH Zurich/SARIT,
service behind a Grid portal. Similarly, are currently investigating how to apply Swizterland
JOpera can be embedded into end-user autonomic techniques in order to extend Tel: +41 1 632 0879
E-mail: pautasso@inf.ethz.ch
Grid-enabled Weka:
A Toolkit for Machine Learning on the Grid
by Rinat Khoussainov, Xin Zuo and Nicholas Kushmerick
In the end of the day, Grids are about Grid-enabled applications. While a number
of general-purpose libraries for scientific computing have been adapted for Grids,
many more opportunities remain in various specialist areas. We describe here
an ongoing work on Grid-enabled Weka, a widely used toolkit for machine learning
and data mining.
Weka is a widely used toolkit for learning (or training) process is to derive tions are performed. On each iteration,
machine learning and data mining origi- a classifier from a set of labelled data (ie one fold is used as a test set, and the rest
nally developed at the University of a set of data instances together with their of the data is used as a training set. A
Waikato in New Zealand. It is a large correct labels). The idea is that a classi- classifier is learned on the training set
collection of state-of-the-art machine fier learned on a labelled data set can and then validated on the test data.
learning algorithms written in Java. then be used to predict class labels for
Weka contains tools for classification, future (unlabelled) data instances. In Grid-enabled Weka, which is being
regression, clustering, association rules, developed in University College Dublin,
visualisation, and data pre-processing. Learning a classifier and using it to label execution of these tasks can be
Weka is open source software under the data can be time consuming and require distributed across several computers in
GNU GPL. It is easily extensible, which significant amounts of memory, espe- an ad-hoc Grid. The labelling function is
allows researchers to contribute new cially for large data sets. Unfortunately, distributed by partitioning the data set,
learning algorithms to Weka, keeping it parallelising and distributing classifier labelling several partitions in parallel on
up-to-date with the latest developments learning is a difficult research problem different available machines, and
in the field. As a result, Weka has on its own. Nonetheless, there are a merging the results into a single labelled
become very popular with academic and number of simpler functions and usage data set. The testing function is
industrial researchers, and is also widely scenarios in Weka that can still benefit distributed in a similar way, with test
used for teaching purposes. from distributing the work on a Grid. statistics being computed in parallel on
The most obvious are labelling, testing, several machines for different subsets of
The main focus of Weka is on classifier and cross-validation functions. the test data. Distributing cross-valida-
algorithms. Simply put, a classifier maps tion is also straightforward: individual
a set of data instances onto a finite set of Labelling involves applying a previously iterations for different folds are executed
classes. Each data instance is described learned classifier to an unlabelled data on different machines. The quality of a
by its attribute values. For example, set to predict instance labels. Testing classifier often depends on various algo-
predicting whether it is going to rain takes a labelled data set, temporarily rithm parameters. The same classifier
based on observations of sky, air temper- removes class labels, applies the classi- may need to be trained with different
ature, humidity, and wind can be viewed fier, and then analyses the quality of the parameters to obtain better results. In our
as a classification task. Each data classification algorithm by comparing system, the user can run a training task
instance includes values of the observa- the actual and the predicted labels. on a remote machine. This allows the
tion attributes, eg (sunny, warm, humid, Finally, for n-fold cross-validation a same user to train several classifiers in
strong), and the available classes are labelled data set is partitioned into n parallel by launching multiple Weka
{rain, dry}. The goal of classifier folds, and n training and testing itera- tasks from the user's computer.
Examples include issues such as nuclear we use quantum mechanical methods, eMinerals minigrid has been developed
waste encapsulation, adsorption of but with smaller numbers of atoms. as an integrated compute and data grid
pollutants on the surfaces of soil parti- with the aim to equip the scientists to
cles, and the interaction of mineral At its heart, the challenges faced in work with more complex simulation
surfaces and fluids. The simulations we scaling up these calculations are compu- studies than before.
perform range in their complexity. In tational, but with larger simulations
some cases we need to simulate systems comes an increased need manage data in The eMinerals minigrid is built around a
containing millions of atoms, and for this new ways. In addition to the fact that heterogeneous set of computing
we represent the forces between atoms larger simulations generate larger data resources, quite deliberately so because
by simply parameterised models. In files, we also have the problem that with different types of molecular simulations
other cases we need to have more accu- higher throughput of calculations comes have very different requirements. The
rate representations of these forces, and the need for intelligent file management compute resources include 4 clusters of
within an individual study. The PCs, an IBM p-series parallel computer,
a Condor pool containing a range of One immediate problem with this regime The final stage in the process is longer-
machines but primarily designed to is that of file transfer. Both Globus and term management of the data. The
allow test jobs with high memory Condor present problems for users in this Daresbury project partners have devel-
requirements, and a large (930 machine) regard. Our solution has been to directly oped a data portal that has a direct inter-
Condor pool containing the teaching PCs incorporate the Storage Resource Broker face with the SRB. Data that require
of University College London. The latter (SRB) developed by the San Diego long-term storage are uploaded to the
is the largest Condor installation in the Supercomputer Centre. The SRB data portal from the SRB, and provided
UK, and is designed to allow high- provides a means to distribute files with metadata that enables use to be
throughput calculations that do not have across a number of file systems (called made of the data by the scientist or
high memory requirements. Access to 'SRB vaults') but to be seen by the user as collaborators at a later date. In order to
these machines is through Globus, with a single file system. The physical loca- ensure interoperability of data, we are
each member of the team having a UK tion of files is transparent to the user, and making use of the Chemical Markup
digital certificate to ensure authorisation is seen only as a file attribute. We have Language, one of the established XML
and authentication. Moreover, we expect set up 5 SRB vaults across the eMinerals standard languages.
the scientists to submit jobs using the minigrid. The integrated compute and
Globus toolkit commands rather than via data grid that constitutes the eMinerals This report represents work in progress.
direct logins to the resources. The only minigrid is shown in Figure 1, with the Components of this work, particularly
logins are permitted to allow code devel- data component shown in more detail in the eMinerals minigrid as an integrated
opers to compile and test. Figure 2. compute and data grid, are now beyond
proof-of-concept and are in production
mode. Immediate tasks are to provide the
tools such as the compute portal to give
Figure 2: the scientists more help in their use of the
The structure of full minigrid. The eMinerals team
the data consists of scientists, applications code
component of developers and computer scientists/grid
the eMinerals specialists from the Universities of
minigrid, Cambridge, Bath, and Reading,
showing how University College London and
the various Birkbeck College, and the CCLRC
components Daresbury Laboratory.
map onto the
Link:
infrastructure of
http://www.eminerals.org
the application
server and Please contact:
database Martin Dove, University of Cambridge, UK
cluster. E-mail: martin@esc.cam.ac.uk
Arteriosclerosis is a widespread disease distributed simulations and visualiza- uses a desktop virtual reality environ-
that manifests particularly in developed tions to support vascular surgeons in ment where the patient’s data, obtained
countries. Treatment often involves making pre-operative decisions. from a scanner, is visualized as a 3D
surgery, such as the placement of stereoscopic image together with the
bypasses that lead the blood around The Virtual Radiology Explorer graphical interpretation of the simulation
clogged arteries to restore normal blood The Virtual Radiology Explorer (VRE), results. A user can then manipulate the
flow. A surgeon plans these interven- developed at the University of 3D images of the arteries. This would be
tions on the basis of 3D images obtained Amsterdam, is a grid-based problem- the virtual surgical procedure, eg the
from MRI or CT scans. Apart from solving environment for virtual vascular addition of a bypass. Blood flow in this
considerations such as accessibility, the surgery. The VRE contains an efficient new geometry is also computed, and the
attainable improvement in the blood mesoscopic computational haemody- user can then assess the effectiveness of
flow will determine what type of treat- namics solver for blood-flow simula- the proposed treatment, and might try
ment is appropriate. Improvements in the tions based on parallel cellular automata. other alternatives to optimise the proce-
support for planning these procedures To convert the medical scans into dure.
are expected to improve their success computational meshes, raw scanner data
rate. We have developed a prototype is first segmented so that only the arterial The medical scanners, the visualization
grid-based system for virtual vascular structures of interest remain in the data environment and the computational
surgery, which may be used during pre- set. The segmented data is then resources required for the flow computa-
operative planning or as a valuable tool converted into a computational mesh. tions and visualizations are usually
in the training of novice vascular The patient’s blood flow is simulated located in disparate geographical loca-
surgeons. The prototype uses advanced using grid resources. The VRE system tions and distinct administrative
domains. The need for transparent access
to these resources, high efficiency and
strict data security triggered the develop-
ment of our advanced grid-based simula-
tion and interactive visualization for
virtual vascular surgery. The middle-
ware to support this was developed in the
CrossGrid project.
CrossGrid
The CrossGrid project is oriented
towards compute- and data-intensive
applications characterized by the interac-
tion with a person in the loop. The
CrossGrid pan-European distributed
testbed shares resources across sixteen
European sites. One key component of
Crossgrid is the Migrating Desktop
(MD) grid portal. The MD produces a
transparent user work environment,
Figure 1: Distributed image-based blood-flow simulation and visualization on the Grid. permitting the user to access grid
From top left to bottom right: A patient is scanned in Leiden, The Netherlands; this resources from remote computers. Users
results in a raw image stored in a Storage Element on the Grid (eg Poznan, Poland); this can run applications, manage data files,
is segmented, filtered and cropped, using a Grid service; a bypass is added; a and store personal settings indepen-
computational mesh is generated (on the local machine) and given to the parallel flow dently of the location or the terminal
solver, running on Compute Elements in the Grid (Amsterdam and Spain); the resulting type. We have incorporated our VRE
flow fields are displayed on the local machine using visualization services offered by system into the Grid via the MD grid
the Grid Visualization Kernel (Linz, Austria). portal. We achieved secured grid access,
Photorealistic Visualisation
with Grid-Based Technology
by Mika Pennanen and Markus Ylikerälä
The rapid development and low price of personal computers make them an
interesting choice for conveying ideas through visualisation. Although real-time
3D graphics are adequate for many applications, a single computer is insufficient
if photorealistic images are required, not only with today’s technology, but even
in the near future. At VTT Information Technology, grid-based distributed
computing has been utilised to take advantage of the processing power of off-
the-shelf office computers that are idle most of the time. As a result, using the
rich resources of this existing distributed and networked environment,
photorealistic images were produced faster and more cheaply.
Over the years, bandwidth has increased has evolved into an accepted way of to the fullest extent, companies have
and become cheaper, computers have connecting computers and computing begun to promote the concept of grid
become more powerful, and the Internet devices. As resources are rarely utilised computing, which enables the integra-
Figure 1:
Ray-tracing
from an
arbitrary
point of view
in an
OpenGL
application.
tion of computing power from anywhere produced more quickly and more processor) varied from nine seconds to
in the world as easily as users now log on cheaply. To obtain photorealistic 47 minutes. With multiple computers,
to the Internet, or even use electricity. images, techniques such as ray-tracing one computer acts as the controller,
This also opens up the possibility, for can be used. Ray-tracing simulates the while the rest perform the actual ray-
example, of increasing the rendering travel of rays of light from the observer’s tracing calculations. As a result, the
capability of low-capacity devices via viewing angle to the light sources. It performance of two computers is worse
the Internet, on demand. Imagine being should be noted that the simulation is than one alone, since control data merely
able to plug your computer into the wall made contrary to physical laws in order increases overhead. Using the increased
and tapping into as much processing to reduce the necessary number of calcu- power of a moderate number of
power as you need. The promise of such lations. This is based on the fact that the computers does produce significant
utility computing is very attractive: it observer can sense only some of the rays. improvements, particularly with large
means obtaining the right amount of The program currently in use is POV- images. However, for large numbers of
computing power at the right time, Ray, a well-known ray-tracing program. computers, the control data overhead
employing the latest hardware, software Unfortunately a single computer – not increases such that the results again dete-
and services, and paying only for the only current machines but those of the riorate. Significant benefits can there-
computing power used. The savings near future – has insufficient power to fore be obtained even with a small
could be substantial. render large or complex photorealistic number of computers. Performance is
images. A grid is therefore used in order not improved by continuously increasing
VTT Information Technology has partic- to gain more computing power. An the number of computers.
ipated in several national and interna- example of our research on ray-tracing
tional research projects related to grid with our OpenGL application is shown Our future work will concentrate on
technology. During these projects we in Figure 1. With this application the extending these solutions and demon-
utilised two of our research areas: utility user can move within a room that has strating them on Grix. Furthermore, it is
computing used for calculation chessboard surfaces, a yellow ball and a planned to extend and implement the
purposes, and visualisation and virtual canvas. The observer takes a snapshot business scenario with payable services
reality environments. The objective was from an arbitrary viewpoint, and a for device users.
to use the grid for visualisation purposes photorealistic image is rendered with a
Please contact:
to convey ideas to the interest group, and grid and attached to the canvas. Mika Pennanen, VTT Information
research has focused on distributed Technology, Finland
calculations for visualisation purposes. To measure the effect of the grid with Tel: +358 9 456 5623
The Globus Toolkit has been used, and a varying numbers of computers, we E-mail: mika.pennanen@vtt.fi
grid called Grix has been developed measured the time it took to render a
Markus Ylikerälä, VTT Information
based on platform-neutral architecture. photorealistic image. The image was Technology, Finland
During our research on distributed calcu- from a model of a ski lodge containing Tel: +358 9 456 6082
lations for visualisation purposes, we two light sources and approximately E-mail: markus.ylikerala@vtt.fi
utilised the processing power of off-the- 55,000 polygons. The complexity of this
shelf office computers that are idle most rendering task varied with the image
of the time. With the rich resources of size; Figure 2 shows the results of these
this existing distributed and networked tasks. Depending on the image size, the
environment, photorealistic images were processing time with one computer (PIII
Grid computing is a recent discipline common in many areas of our lives: computing, data-intensive computing
related to the utilization of large-scale telecommunications, economy, bio- and collaborative computing. At this
distributed systems for a given purpose informatics, industrial environments,and stage of our research we are primary
by taking advantage of the rich infras- so on. For such a wide spectrum of prob- involved with the two first topics, ie
tructure provided by the Internet. To lems, heuristics come to the rescue, since distributed and high-throughput
understand the basic idea that motivates exact techniques are unable to locate any computing. However, we plan to enter
our research, consider a grid of kind of solution. This area is known as the other three domains by developing an
computers as a huge virtual multi- Networking and Emerging Optimization open optimization service for the
computer ready for processing, storage (NEO), and the GISUM group at the Internet, solving data-mining problems
and communication. Since a grid can be University of Málaga (Spain) is working and facing software agent applications
made up of a set of geographically sepa- on just these techniques. respectively. Figure 1 is an example of
rate networks, enormous computer the goal of the present work: computing
power is available for solving complex In short, using Grid technology, our aim an exact Pareto front.
problems that are limited in CPU and is to solve optimization problems that are
that require long delays if solved in otherwise out of the scope of researchers Multi-criterion optimization in which
modern computer LANs. dealing with parallel algorithms. Five several (non-dominated) optimum solu-
types of applications are usually identi- tions must be found is a promising
Complex problems that can only be fied as related to Grid computing: research field. At present, this field lacks
solved in non-polynomial time arise in distributed supercomputing, high- algorithms that could ensure the compu-
most fields of research and are becoming throughput computing, on-demand tation of the exact Pareto front for a
general problem. We avoid this inconve-
nience by using an enumerative-like
search that computes all the non-domi-
nated solutions in a grid. Later,
researchers can use these results to find
efficient heuristics achieving similar
(optimal) results. We are in the first stage
of research, and have solved problems at
Málaga using a modest grid of around
110 processors. As simple as it sounds,
finding the optimum Pareto front is
extremely hard, even for small problems,
Figure 1: A simple example of a multi-objective optimization problem with the and with most algorithms it is not guar-
constraints (left), and its Pareto front (right). anteed that the optimum front will be
located for arbitrary problems as we do
with our grid exact algorithm.
because users must deal with certificates, (http://tracer.lcc.uma.es), which is devel- developed. We are currently testing new
installation details and production tools oping advanced algorithms in close grid algorithms derived from well-
that require a time-consuming learning connection with the Internet. Two known standard techniques like PAES or
curve. However, Globus actually eases successful case studies in TRACER NSGA II.
the next step of our research in relating to Grid computing can be cited.
connecting to foreign grids. In addition, Firstly, a grid algorithm was developed We invite everyone to look at our results
we have tested the network usage and to compute the exact Pareto front of and contact us with any comments or
execution times of heuristic algorithms several hard multi-objective problems by suggestions for collaboration.
and tools like MPI on Globus with very using Condor in a grid with more than
Link:
satisfactory results, and this suggests that 110 computers http://neo.lcc.uma.es
we could achieve fruitful research with (http://neo.lcc.uma.es/Software/ESaM).
Globus in the future. At present, we are extending the grid to Please contact:
other sites in Spain and Europe. Enrique Alba, University of Málaga/
Most of this work has been undertaken Secondly, grid-aware multi-objective SpaRCIM, Spain
Tel: +34 952 132803
within the on-going TRACER project heuristics, initially run on Globus, were E-mail: eat@lcc.uma.es
DB-Annotator was conceived for the The project DB-Annotator (see Figure 2) residing in RDBs can be derived from the
Resource Description Framework (RDF) was designed for universal annotation of tables (entities) themselves, and from the
annotation of structured information data that can be represented in tabular columns (entity attributes) and the cell
sources such as relational data or XML- form, ie data that can be qualified by contents (instances of attributes). Still
based service descriptions. Only semanti- unique keys. more fine-grained information comes
cally annotated Grid services (GS) from the notorious ‘description’ fields,
provide a means of finding data or The main thrust for the development but putting this implicit information to
compute-services that are suitable for the came from the realization that most inter- use for machine-reasoning is unrealistic
task at hand. Furthermore, they make the esting data within biological databases at this point.
building of workflows (coupled services resides within free-form text fields and is
within the grid) a realistic goal. DB- therefore not easily accesible for data The goal was thus to provide easy navi-
Annotator will support several levels of mining. Relational databases (RDB) are gation through existing ontologies and
semantic annotation to enrich Grid usually the only choice for data retrieval data-source-independent RDF annotation
services, ranging from the services them- within distributed organizations. of RDB data via drag and drop (see
selves to the data within the Grid. Categorial description of the typed data Figure 2). This data categorization can
take place on three levels within a like ‘entity_xyz is-a category_abc’), its The project was launched in the depart-
database (see Figure 3), not counting the output can also serve to enhance UDDI, ment for bioinformatics at Fraunhofer
database itself. A unique data key that a proposal made for the introduction of SCAI. It is a collaborative project
can be linked to an ontological category content-searchable Web services and a between the bioinformatics department
will therefore appear as possible starting point for GS-workflow and the department for Web applica-
‘database:schema:table[:column:row]’, construction. Instead of the database tions. Preliminary work commenced late
the latter two key parts being optional unique keys, the RDF would contain in 2003 with the aim of producing a
depending on the depth of the annotation. qualifiers for the available Grid services. prototype that would capture the basic
Eventually a semantically enhanced Grid functionality of the future system. The
This concept of assigning ontology Service-Oriented Architecture (gSOA – prototype, featuring 1:n mapping of a
classes to data was inspired by existing SOA being a term borrowed from the database to several ontologies and RDF
annotation software for the Semantic Web services world) would look like flatfile storage, was completed along
Web. Once a data source has received Figure 1. Full categorization of both GS with a detailed requirement analysis this
ample annotation, data mining will be and the exchanged data leaves no more summer. A new structured development
possible by employing standard methods room for ambiguities – a precondition from scratch has now started and is
on the categorized data. With annotated for data exchange and manipulation expected to be finished early in 2005.
RDB, access to data services in the Grid within the Grid network of virtual orga- Among the things that will be included
will improve significantly, since querying nizations (VO). in the new phase are n:m relationships
by categories is enabled across several between data and ontology classes and
RDBs at once. Further, every result set Given the bulk of biochemical entities storage in a single, structured relational
will be enhanced with the respective cate- that go under various names and interact repository. The product is currently
gory information, a data basis that could on different levels, the relevance of fully intended to provide support for semantic
be used to support further automated annotated data Grid services in bioinfor- annotation of biological and chemical
annotation of newly inserted data. matics, especially systems biology, is databases. The possibilities for Grid
enormous.. Until now, ambiguities in services annotation within UDDI are
Since DB-Annotator produces standard synonyms and interactions had to be explored within the semantic grid
RDF (subject-predicate-object triples detected by human intelligence,– research group under the auspices of the
meaning full-fledged Global Grid Forum (GGF).
systems biology
Links:
services won't be Globus Toolkit: http://www.globus.org
successfully auto-
Semantic Web Services:
mated without proper http://swws.semanticweb.org
categorization. Since
Global Grid Forum: http://www.gridforum.org
DB-Annotator draws
on a common set of Gene Ontology: http://www.geneontology.org
agreed ontologies (eg Please contact:
Figure 3: Selection modes in DB-Annotator. Data entities for gene ontology) for the Kai Kumpf, Institute for Algorithms and
annotation can either be the whole table, one column or one annotation of both data Scientific Computing (SCAI), Fraunhofer ICT
cell. The data key for annotation is constructed depending services and data sets Group, Germany
on scope of the selection. Only (multiple) annotation of a or subsets, it can Tel: +49 2241 14 2257
E-mail: kai.kumpf@scai.fhg.de
single cell corresponds to an extension of the data model. provide the semantic
Selection of a row corresponds to selection of a record from glue on all levels
the extended data model on the basis of table and ontology required.
keys and as such need not be considered separately.
The CoreGRID Network of Excellence activity is structured around six comple- tance to ensure a durable development
(NoE) aims at strengthening and mentary research areas that have been and deployment of GRID infrastruc-
advancing scientific and technological selected on the basis of their strategic tures:
excellence in the area of Grid and Peer- importance, research challenges and • Institute on Knowledge and Data
to-Peer technologies. It is funded by the existing European expertise to develop Management
European Commission through a grant next-generation grid middleware. • Institute on Programming Model
of 8.2 million Euro for a duration of four • Institute on System Architecture
years. To achieve its objective, the Virtual Institutes • Institute on Grid Information and
Network brings together a critical mass To ensure progressive evolution and Monitoring Services (WP5)
of well-established researchers (119 durable integration, the Network is oper- • Institute on Resource Management
permanent researchers and 165 PhD ated as a European Research Laboratory and Scheduling
students) from 42 institutions — several — known as the CoreGRID Research • Institute on Problem Solving
of which belong to ERCIM — who have Laboratory — having six institutes. Each Environment, Tools and GRID
constructed an ambitious joint program of them is dedicated to the particular Systems.
of activities. This joint program of domain identified as of strategic impor-
The Network is thus committed to
setting up this Laboratory and making it
internationally recognised and sustain-
able. It is funded by a European grant
that is assigned to the CoreGRID NoE
for a duration of four years to cover the
integration costs, while the network part-
VTT ners cover the expense required to
IC
RAL-CCLRC perform the research associated with the
U. Belfast joint program of activities. Integration is
U. Cambridge KTH achieved by the joint execution of
U. Cardiff SICS
U. Manchester research projects operated through
U. Newcastle FhG research groups within the six institutes,
U. Wesminster FZ-Julich the sharing of a common grid testbed for
Delft U. HLRS research assessment, the access to a
Vrije U. U. Munster PSC
U. Chile U. common communication infrastructure
CETI CYFRONET
Dortmund to ensure collaboration and dissemina-
C U. Passau
UCL ZIB Masaryk U. tion, and a coherent management frame-
INRIA EPFL SZTAKI work to encourage mobility of senior and
CNRS
ERCIM
UASF post-doctoral researchers and PhD
students. In addition to these classical
U. Coimbra CLPP-BAS
UPC CNR-ISTI integration activities, the network has a
INFN proactive role to increase the awareness
U. Calabria
U. Lecce
ICS-FORTH of trust and security technologies among
U. Cyprus
U. Pisa the network participants.
pants will be of mutual benefit to the organized to ensure that CoreGRID is a and the establishment of a repository of
European Grid community (both leading educational source in Europe in reference implementations and grid
academic and industrial), CoreGRID Grid and Peer-to-Peer technologies. A middleware.
will organize a set of activities to spread dedicated activity under the form of an
Link:
excellence outside the Network. A Industrial Advisory Board will target http://www.coregrid.net
public Web site is an essential mecha- mainly industry and commerce to ensure
nism for the international Grid commu- a strong interaction and involvement. Please contact:
nity to present the Network as a Bruno Le Dantec, ERCIM
Research Laboratory. A set of Finally, CoreGRID is committed to Administrative & Financial Coordinator
Tel: +33 4 92 38 50 13
CoreGRID publications can be accessed participating in collaboration activities E-mail: Bruno.Le_Dantec@ercim.org
thanks to this Web site (technical that can benefit the other EC Grid
Thierry Priol, INRIA, France
reports, network brochure and newslet- projects. In particular, it will take on a Scientific Coordinator
ters). The Network will organize a set of leadership role in the organization of Tel: +33 2 99 84 72 10
thematic workshops and conferences on training activities and will collaborate on E-mail: Thierry.Priol@inria.fr
an annual basis. Training activities are EU grid roadmaps, research inventories
GridCoord – Co-ordinating
Europe-wide Initiatives in Grid Research
by Marco Vanneschi
There are several Grid research initia- order to achieve critical mass, avoid ment by a number of Member and
tives, ongoing or planned, at the national duplication and reduce fragmentation of Accession States (UK, France, Italy, The
and European Community level, aiming efforts. Netherlands, Germany, Hungary, Spain,
at the development of a rich set of Poland, Czech Republic, Sweden) is
advanced technologies, methodologies From 2002 to 2006, EU funding for Grid about 300-500 M .
and applications. If the challenges that research and deployment more than
lie ahead are to be solved, enhanced doubled, reaching 275 M in FP6. For The totality of these initiatives could
coordination among the authorities the same period, an estimate of the provide the EU with the potential to play
funding these initiatives is needed in funding for Grid research and deploy- a leading world role in Grid technologies
and applications. National and EU • Germany: University of Stuttgart, To achieve the third objective — devel-
collaborations have been established High Performance Computing Center opment of Grid R&D programmes and
with other international players (in the (HLRS), and Konrad-Zuse-Zentrum road maps — the project will:
US and Asian-Pacific area) and with für Informationstechnik (ZIB), Berlin • develop a coherent and innovative
international standards organisations. • The Netherlands: University of Grid R&D programme vision in
However, if Europe wishes to compete Amsterdam, Faculty of Science, Europe taking into account cohesion,
with leading global players, it would be Institute of Informatics and considering present achievements
sensible to attempt to better coordinate • Hungary: Computer and Automation and the multinational character of Grid
its various, fragmented efforts in order to Research Institute, Hungarian R&D in Europe. This task will address
achieve critical mass and the potential Academy of Sciences MTA SZTAKI the strategic research issues identified
for a more visible impact at an interna- • Spain: Technical University of by the EU Expert Group on Next
tional level. This requires: Madrid, School of Computer Science; Generation Grids (June 2003) and the
• co-ordination among the funding • Poland: Poznan Supercomputing and UK Gap Analysis report (May 2003),
authorities in order to maximise the Networking Center which have been substantially stimu-
effective use of the considerable • Sweden: Swedish Agency for lated by several ongoing national and
existing and potential resources Innovation Systems (VINNOVA). European research initiatives and/or
• collaboration among individual collaborations at international level
researchers and the creation of Objectives and Activities • develop Grid R&D road maps and
European excellence and competence To achieve the first objective — implementation strategies on the basis
centres improved coordination among funding of the Grid R&D programme vision.
• a visionary research agenda. bodies, policy makers and leaders of
Grid initiatives — the project will: The multinational character of Grid
The first objective of GridCoord is thus • develop a comprehensive research in Europe should be maintained
to strengthen co-operation among the compendium of national and EU and developed according to the intrinsic
funding authorities in order to better co- initiatives including an analysis of strengths of the different national
ordinate the planning of future activities strengths, weaknesses and gaps at EU programmes. This implies the continued
in the field of Grid research, an ERA level development of both individual national
Strategic objective. A second objective • identify common interests and syner- and EU programmes for future Grid
is to enhance already existing collabora- gies capable of leading to cooperation research initiatives at the same time as
tions between researchers and users. between different groups, collabora- supporting multilateral exchange of expe-
Finally, it is our intention to contribute tion among funding authorities and riences and plans to create a more
towards the development of visionary eventually the planning of common complementary and co-ordinated
national and EU programmes and initiatives to overcome fragmentation, approach towards Grid Research in
roadmaps, enabling Europe to play a avoid duplication of efforts and build Europe.
leadership role in Grid technologies and the necessary critical mass
applications. • provide a regular forum at which GridCoord is coordinated by the
participants from national funding Dipartimento di Informatica, University
Partners authorities and leaders of national Grid of Pisa, Italy; Coordinator: Marco
The project brings together national initiatives can meet at six-monthly Vanneschi; Project Manager: Roman
research programme leaders and repre- intervals to establish synergy between Tirler. The starting date was July 2004,
sentatives from national funding authori- the different projects and programmes. duration 18 months, budget 960,000 ,
ties with the aim of ensuring national for a total cost of 1155 .
commitment to future joint activities. To achieve the second objective —
Link:
The following institutions are partners of enhanced collaboration among http://www.gridcoord.org
GridCoord: researchers and users of Grid technology
• Italy: University of Pisa, Department — the project will: Please contact:
of Computer Science, and University • organise general and specialized Marco Vanneschi, Dept. of Computer
of Genova, Department of workshops Science, Pisa University, Italy
Tel: +39 050 2212738
Communication, Computer and • promote and organise E-mail: vannesch@di.unipi.it
System Science dissemination/education/training
• France: Institut National de Recherche initiatives both for users and for
en Informatique et Automatique research centres
(INRIA), Sophia Antipolis, and • investigate the status of standard-
University of Nice, Sophia Antipolis isation activities in Grid computing
• UK: Engineering and Physical and devise a strategy for European
Sciences Research Council (EPSRC), activities in standardisation.
and Queen's University Belfast,
School of Computer Science
Recent years have seen an unprece- spanning national and enterprise Emerging ICT paradigms such as
dented acceleration in the evolution of borders, where the participating entities Autonomic computing, Utility
the Internet as the technological vehicle (enterprises or individuals) pool computing and Grid computing are
underpinning the expansion of service resources, information and knowledge in making the formation and operation of
provision and inter-/intra- enterprise order to achieve common objectives. virtual organisations easier by providing
integration in all market sectors. This The objectives may be short term - eg to dynamic management of the distribution
brings about the prospect of ad hoc inte- deliver a one-off service in response to a of computational processes across avail-
gration of systems across organisational specific customer demand - or long- able resources. Notwithstanding the
boundaries to support collaborations that lasting. In the latter case, the VO's struc- major ICT breakthroughs of the last two
may last for a single transaction or ture, business processes and operational decades, protecting one’s assets while
evolve dynamically over many years. infrastructure must adapt as the goals of integrating services, processes and
This sets new requirements for scala- the collaboration, the participating enti- resources, remains a major ICT chal-
bility, responsiveness and adaptability ties, the business context and the tech- lenge. Overcoming such challenges
that necessitate the on-demand creation nologies employed, change. requires the development of novel tech-
and self-management of dynamically nology realising innovative ideas over
evolving virtual organisations (VO) widely acceptable interoperable plat-
forms. The required scalability, respon-
siveness and adaptability for dynamic
virtual organisations, makes the provi-
sion of cost effective trust and contract
management solutions for VO environ-
ments, the most demanding and timely
research challenge in this field. Effective
solutions require interdisciplinary
approaches, at a global level, integrating
tools from law, cognitive and social
science in addition to telecommunica-
tions and computing.
TrustCOM Consortium.
• a novel trust and contract management security, trust and contract manage- The TrustCOM Consortium
reference architecture that will enable ment based on Web and Grid services The TrustCoM consortium brings
collaborative work within on-demand technology together researcher from many major
created and self-managed dynamic • system and software engineering tools middleware vendors and provides a
VOs leveraging on the emerging and methods analysis the VO life- balanced blend of academic and applied
convergence of Web Services and cycle and offering a library of design researchers, end-user organisations, and
Grid technologies patterns and generic software compo- enterprises looking to utilise results in
• a set of conceptual models explaining nents implementing selected services products and services. As such it is well-
the fundamental concepts, principles that offer the core functionalities of the placed to define, conduct and exploit
and methods underpinning the above VO leading edge research that is relevant to
architecture. Effectively these provide • testbeds exhibiting instantiations of the needs of European business, govern-
the meta-model of any new architec- the above architecture and reference ment and society.
tural constructs that may result from implementation into two classes of
Link:
TrustCoM research realistic application scenarios, namely http://www.eu-TrustCoM.com
• a set of profiles, that bring together collaborative engineering (CE) and
and potentially extend selected provision of ad-hoc aggregated Please contact:
Web/Grid Services specifications at services (ADP) Theo Dimitrakos, Scientific Coordinator,
specific version levels, along with • selected demonstrators exhibiting the CCLRC, UK
E-mail: T.Dimitrakos@rl.ac.uk
conventions about how they work business value and benefits of the
together to support potential imple- TrustCoM framework in the above- Michael Wilson, Programme Manager,
CCLRC, UK
mentations of the TrustCoM frame- mentioned application domains E-mail: M.D.Wilson@rl.ac.uk
work • studies analysing selected aspects of
Santi Risto, Operations Manager, Cataluña
• a reference implementation of the the legal and socio-economic context Atos Origin,Spain
above integrating and extending that underpins such Virtual E-mail: Santi.Ristol@atosorigin.com
already established or emerging inter- Organisations.
operability standards for autonomic
Grid.it is a strategic project in the area of Enabling Technologies for the Information
Society, coordinated by the Italian National Research Council. The project has a
strong interdisciplinary character and is aimed at defining, implementing and
applying innovative solutions for networked computing-enabling platforms,
oriented towards scalable virtual organisations and based on the 'Grid Computing'
paradigm. The research topics in the project span from high performance photonic
networks to innovative middleware services, high-performance programming
environments and so forth.
Moving on from the distributed platform extends to all the platform levels, from parallelism. The programming environ-
based on a distributed infrastructure high bandwidth networks to middleware ment must therefore be characterized by
scenario, Grid.it places special emphasis services and, in particular, to resource a high degree of portability on different
on the high performance requirements of management as well as tools and hardware-software systems (or different
applications developed on the Grid. This programming environments. hardware-software combinations) in a
means that the integration of systems and heterogeneous and dynamic context.
resources, as well as heterogeneity and At the programming tools and environ- Portability must be guaranteed not only
dynamic situations management, must ment level the high performance require- for code but should also ensure that the
explicitly handle the case of Grid nodes ment implies that, when designing scal- performance matches the configuration
(which in general are geographically able virtual organisations (VOs), a of the target system at hand. Tool inter-
distributed or located on private virtual unifying approach should be used. This operability and high performance appli-
networks) as high performance systems, approach takes into account both aspects cation reuse are also fundamental to the
such as parallel machine architectures or related to the distribution of computa- programming environment.
clusters. Research on high performance tions and resources, and those related to
At the middleware level, research on • Grid portals: Grid enabled application Organisation
resource management includes aspects services, eg providing the user with the Six Italian research units (ISTI-CNR ,
of maximum importance such as possibility to submit tasks to remote ISTM-CNR, ICAR-CNR, INFN, CNIT,
discovery, brokering, scheduling, moni- jobs and collect results via Web inter- ASI) participate in the project and also
toring and performance evaluation/ faces. involve a large number of Italian univer-
prediction. sity departments.
The design and implementation of scien-
The study of the high speed networks tific libraries that can be used in a hetero- The principal investigator is Marco
needed to support enabling Grid plat- geneous and dynamic context, such as Vanneschi, Pisa University and ISTI-
forms for scalable VOs is an internatio- the Grid, completes the research on CNR, Pisa. The Technical Board is coor-
nally recognized 'hot' topic. Within this programming environments. dinated by Domenico Laforenza, ISTI-
research activity an important role is CNR, Pisa.
played by experiments on very high Applications
bandwidth optical networks, based on Grid.it includes the development of The Grid.it project aims at playing an
photonic technology for Grid platforms some demonstrators selected within important role in the training of highly
with high performance sites that extend application fields that are of maximum qualified young people. A relevant part
to the metropolitan area. interest, not only for their scientific of the project budget is reserved for
value, but also as testbeds for high contracts for young researchers.
In addition to aspects concerning performance Grid platforms:
programming environments and • earth observation The project has funding of 8.1 M from
resource management, the software tech- • geophysics the Italian Ministry for Education and
nology studied in Grid.it includes some • astronomy Research for three years (November
fundamental aspects related to middle- • biology 2002 – November 2005).
ware: • computational chemistry.
Link:
• security: secure Grid environments http://www.grid.it
and cooperation among Grid environ- In order to be able to implement and
ments belonging to different organiza- experiment the ideas and results of the Please contact:
tions project, a Grid infrastructure will be Domenico Laforenza, ISTI-CNR, Italy
• data intensive services: federated implemented on a national scale, based Tel: +39 050 3152992
E-mail: domenico.laforenza@isti.cnr.it
database services, visualization and on the GARR network and using, in
hierarchical management of data and some metropolitan sites, optical fiber Marco Vanneschi, Pisa University, Italy
Tel: +39 050 2212738
meta-data according to advances and interconnections. This project result will E-mail: vannesch@di.unipi.it
high-performance techniques provide the community with a national
• knowledge discovery services: Grid Grid that can be used in different compu-
services (data mining, search engines, tational science research sectors, and
etc.) that provide consistent, efficient also for commerce, industrial and social
and pervasive access to high end service applications.
computational resources
Multimedia is everywhere. Libraries and Context Dependency tion system should be able to handle
museums digitize their collections and Yet high-quality extraction algorithms these ambiguities.
make them available to interested alone are not enough. Equally important
parties. Cheap disk space has brought the is a system to coordinate the extraction Valid Sentences
storage of large amounts of multimedia of annotations. The main problem is that CWI solves these difficulties by
within everyone’s reach. At the same annotations depend on context, and in combining database technology with
time however, it has complicated the practice this means they depend on each ideas from formal language theory to
retrieval of objects. other. This has consequences for the way form the Acoi annotation management
extraction algorithms should be used. system. Its basis is the theory of feature
Not only have databases become larger, For example, feeding charts or logos to a grammar systems. Such a system not
they also contain more types of media. face extractor would be a waste of only describes the annotations them-
While most people are able to find a text system resources and time. The annota- selves, but also their dependencies and
containing certain keywords, anyone tion system makes sure the extractor is contexts. Feature grammar can be
who has ever used Google Image Search only applied to images with a high compared to grammar in natural
to look for a specific picture knows that chance of containing faces, like photos language. Grammar determines which
finding what you need can be far from or drawings. word classes can be combined and in
easy. The majority of search engines can
only interpret textual data. They cannot With the advent of digital
‘see’ what is depicted in an image or cameras, DVDs and MP3s it
‘hear’ what is on an MP3. has become very easy to
compile large multimedia
Annotation libraries. At the same time,
One way of dealing with this problem is it has become increasingly
to annotate media objects in advance. difficult to effectively
Annotations describe particular features search collections that
of the stored media objects and can be contain a wide range of
used to guide semantic search queries. media types like text,
For example, when MP3s are annotated images, movies and audio.
for genre and background information
on the performing artist, search queries
like ‘find me all blues songs played by A lack of context also complicates incre- what order to form a valid sentence. A
guitarists from Mississippi’ can give mental maintenance. Without know- feature grammar system does the same
meaningful results. ledge of context and interdependencies, for annotations and extraction algo-
the entire annotation process must be rithms. It determines what extraction
Annotation can be done manually, but rerun every time an extraction algorithm algorithms must be called to form a valid
for large multimedia collections this is added or replaced. The annotation annotation ‘sentence’.
quickly becomes impossible. It is there- system finds dependencies and deter-
fore necessary to turn to automatic anno- mines exactly which annotations should Since the feature grammar system also
tation using extraction algorithms. These be updated and which can be re-used. stores each annotation’s place in the
algorithms can perform easy tasks, such network of interdependencies, incre-
as finding the length of an MP3, or Context dependency of annotations can mental maintenance is possible. When
complicated ones, like detecting human also cause extraction algorithms to give updating the database, sentences can be
faces in images. Designing clever extrac- ambiguous answers, especially when reinterpreted to determine which extrac-
tion algorithms is a necessary condition dealing with complex features. This can tions must be redone. Furthermore, tech-
for an automatic annotation system. result in outputs like ‘this image either niques from formal language theory
contains a human or a pig’. The annota- could be modified and used in the anno-
tation system. Resolving ambiguities, for
Traditionally, use cases are written in plain English to make the requirement
specifications accessible to a wide audience. It is widely accepted that automatic
processing of textual use cases is not feasible in general; this is considered as
a trade-off for the high readability of textual use cases. Nonetheless, there is
ongoing research exploring the options to apply linguistic techniques to textual
use cases. The Procasor project of the Distributed Systems Research Group at
the Charles University in Prague aims to convert textual use cases into event-
based behavior specifications, behavior protocols.
The functional requirements for a future natural language. The guidelines for vs. an actor), it is possible to determine
system are typically documented with writing use cases significantly restrict whether the action described by the step
use cases. In the traditional approach, the language by asking for simple and is receiving a request, sending a request,
plain English is used for use cases. A uniform sentence structure; a use case or an internal action of SuD.
strong benefit of this approach is that use sentence should be in active voice,
cases are comprehendible to a wide audi- following the simple pattern 'subject – We illustrate the approach on the parse
ence. A textual use case describes the verb – direct object – prepositional tree of step 1 shown in Figure 2. The
possible scenarios of the future system, phrase'. The subject should be the entity subject of the sentence, 'Seller', a noun
by showing how a particular goal is (SuD or an actor) active in the task. (NN), is the name of an actor involved in
achieved by the System under design the use case. Therefore, the step
(SuD) and its surrounding actors in a Readily available linguistic tools, eg, the describes an event (a request) received
sequence of steps. A step of a textual use statistical parser developed by Michael by SuD from the entity Seller.
case describes an action that is either a Collins at the University of Subsequently, we estimate the event
request or information passed between Pennsylvania, permit obtaining the parse label for representing the request. In the
an actor and SuD actors, or an action tree for a sentence. While the parsing is parse tree, we identify the verb (VBZ)
internally processed by SuD. Figure 1 not 100% reliable (it is statistical 'submits' and the direct object 'item
shows a fragment of a textual use case. parsing), the accuracy improves for description'. Depending on conceptual
simple sentences – as is the case for use objects captured in the domain model,
An obvious drawback of writing use cases. The simple sentence structure the words constituting the direct object
cases in plain English is the lack of permits us to obtain the principal phrase are selected; here we acquire the
formal tools to reason upon the require- attributes of the action described by the event token ?SL.submitItemDescription
ment specifications. Also, the subse- sentence from its parse tree. With only a to represent the step 1. The step 2
quent stages of the software develop- simple domain model (names of entities describes an action internally performed
ment process have to start with manually and a list of conceptual objects), we can by the SuD, which we represent with the
processing the use cases – eg, identifying identify the entity active in a particular event token #validateDescription.
operations of future objects. Luckily, use step of a use case. Depending on the role
cases are typically not written in plain of the active entity in the use case (SuD
Figure 1: Fragment of a textual use case. Figure 2: Parse tree of sentence "Seller submits item description".
In a subsequent stage, we convert the tical natural language parser, the output is behavior protocols to detect inconsisten-
whole use case into a behavior protocol only an estimate of the behavior specifi- cies in use case specifications.
(IEEE Trans. Software Eng. 28(11), cation; nonetheless, it can prove useful in
Link:
developed within our group), a formal obtaining an initial design of interfaces of http://nenya.ms.mff.cuni.cz/
method featuring, eg, a decidable compli- the future system. Our future work
ance relation. We have implemented the focuses on enhancing the transformation Please contact:
conversion outlined here in the Procasor to produce matching event tokens for Vladimir Mencl, Charles University Prague/
tool (Pro-case stands for Protocol use complementary actions; the long-term CRCIM,Czech Republic
Tel: +420 2 2191 4232
case). As the tool is employing a statis- goal is to apply reasoning available for E-mail: vladimir.mencl@mff.cuni.cz
GHT* is a scalable and distributed similarity search structure that has been
specifically designed to support metric space objects. The structure is based on
the Peer-to-Peer (P2P) communication paradigm. It is scalable, which means that
the query execution achieves practically constant search time for similarity range
queries in data-sets of arbitrary size. Updates are performed locally and node
splitting does not require sending multiple messages to many peers.
P2P communication paradigms are vector space is covered by a metric space functions to decide the closeness of
quickly gaining in popularity due to their with an appropriate distance function objects as a distance, which can be seen
scalability and self-organizing nature, (for example the Euclidean distance), as a measure of the objects 'dis-simi-
forming the basis for building large- even n-dimensional vector spaces are larity'. For any distinct metric objects x,
scale similarity search indexes at low handled easily. Furthermore, there are y, z, the distance must satisfy the proper-
cost. However, most of the numerous numerous metric functions able to quan- ties of reflexivity, d(x,x) = 0, strict posi-
P2P search techniques proposed in tify similarity between complex objects, tiveness, d(x,y) > 0, symmetry, d(x,y) =
recent years have focused on single-key such as free text or multi-media object d(y,x), and triangle inequality, d(x, y) ≤
retrieval. A good example is the Content features that are very difficult to manage. d(x, z) + d(z, y).
Addressable Network (CAN), which is a For example, consider the edit distance
distributed hash table abstraction over defined for sequences and trees, the The distributed environment is
Cartesian space. Hausdorff distance applied for composed of network nodes (peers),
comparing shapes, or the Jacard coeffi- which hold metric objects, execute simi-
Our objective is to develop a distributed cient, which is often used to assess the larity queries on stored data and commu-
storage structure for similarity search in similarity of sets. Much research is now nicate with other peers. Every peer is
metric spaces that would scale up with focused on developing techniques to uniquely denoted by its identifier PID.
(nearly) constant search time. The structure collections of metric objects so Peers hold data in a set of buckets. Each
advantage of the metric space approach that search requests can be performed bucket has a unique identifier within a
for data searching is its 'extensibility', efficiently. peer, designated as BID. Each peer also
allowing us to perform exact match, maintains a tree structure called the
range, and similarity queries on any A convenient way to assess similarity address search tree (AST). This is used
collection of metric objects. Since any between two objects is to apply metric to route similarity range queries through
the distributed network. AST is based on The most important advantage of GHT*
the metric data partitioning principle with respect to single-site access struc-
known in the literature as the tures is its scalability through paral-
Generalized Hyperplane Tree (GHT). lelism. As the size of a data-set grows,
GHT is a binary tree with metric objects new server nodes are plugged in and
stored in leaf nodes implemented as both their storage and the computing
buckets of fixed capacity. Inner tree capacity are exploited. Figure 2 shows
nodes contain selected pairs of objects the result of the parallel response time of
called pivots. Respecting the metric, the a range query, determined as the
objects closer to the first pivot appear in maximum of the costs incurred on
the left subtree and those closer to the servers involved in the query plus the
second are in the right subtree. search costs of the AST. For evaluation
purposes, we use the number of distance
We start by building a tree with only one computations (both in the AST and in all
root node represented by a bucket B 0 the accessed buckets) as the computa-
When the bucket B0 is full we must split tional costs of a query execution. This
it: we create a new empty bucket B1 and experiment shows how the parallel
move some objects (half of them if Figure 1: Split of a GHT bucket. search time becomes practically constant
possible) into it to gain space in B0. See for arbitrary data volume and the larger
Figure 1. The split is done by choosing a the dataset the higher the potential for
pair of pivots p 1 and p 2 from B 0 and interquery parallelism. This is the most
moving into bucket B1 all objects o that important feature of GHT*.
satisfy the condition d(p1, o) > d(p2, o).
Our future work will concentrate on
Pivots p1 and p2 are associated with a strategies for updates (object deletion)
new root node and thus the tree grows and pre-splitting policies, plus more
one level. This split algorithm can be sophisticated mechanisms for organizing
applied on any leaf node and is an buckets.
autonomous operation (no other tree
Please contact:
nodes need to be modified). Claudio Gennaro, ISTI-CNR, Italy
Figure 2: Tel: +39 050 315 2888
Parallel response time of a range query. E-mail: claudio.gennaro@isti.cnr.it
The Centre for Mathematics and Computer Science (CWI) performs a wide range
of research. Consequently, its high-performance computing facilities must meet
diverse requirements. This has led to the purchase of a cluster with an asymmetric
core-edge design. Under the influence of this asymmetric structure, some cluster
applications are evolving into Grid applications.
Many research groups at CWI share a edge serves as a compute farm. It The Database Architectures and
need for high-performance computer contains several groups of machines, Information Access group works on
facilities. In the FACS (Facilitating the whose main task is to provide CPU database technology and needs machines
Advancement of Computational cycles. The core consists of a small with both as much memory as possible
Science) project, three of these groups number of machines, connected by a and huge, fast storage. For this type of
combined their resources to set up a high-bandwidth low-latency network work, the bandwidth between memory
large shared cluster. Since each user (Infiniband) and each equipped with a and storage is critical. Furthermore, part
group has its own specific demands on large amount of memory. Currently, we of the research consists of designing
memory, CPU power and bandwidth, the have 22 (mostly dual CPU) machines in highly efficient systems for extremely
FACS cluster deviates from the more the edge and two quad AMD Opterons large databases. Since the research
usual symmetric design. The cluster's with 16G each in the core. considers typical database hardware set-
INRIA copyright at
http://www.syndex.org.
Large-Scale Simulations
of Diffusion in Cell Biology
by Ivo F. Sbalzarini, Anna Mezzacasa, Ari Helenius
and Petros Koumoutsakos
Due to the high structural organization of We propose the deterministic particle ical structures using the geometry of real
the intracellular space, diffusion is method of Particle Strength Exchange samples.
always restricted to the specific shape of (PSE) for simulations of diffusion in
the organelle under consideration. To cellular geometries of realistic Due to the complex shape of the compu-
determine and quantify the diffusive complexity. PSE was introduced as an tational domain, a large number of parti-
mobility of substances within live cells alternative to the method of random cles is usually needed to resolve the
the technique of fluorescence recovery walk, enabling higher-order simulations concentration field. We implemented an
after photobleaching (FRAP) is widely of convection-diffusion processes. PSE efficient parallel PSE code based on a
employed. is a grid-free, deterministic particle general-purpose Parallel Particle Mesh
method and thus combines the advan- (PPM) Library currently under develop-
From a practical point of view a method tages of high-order convergence and ment. The code provides different
is needed to deduce molecular (micro- geometric flexibility. It enables efficient domain decomposition techniques,
scopic) solute diffusion constants from simulations of solute diffusion in biolog- dynamic load balancing among inhomo-
measured fluorescence recovery data. geneous processor clusters, parallel file
With the establishment of the IEEE- become more demanding in terms of and/or transmit power. Three main
802.11a/b/g standards, wireless throughput and link reliability (quality of effects are responsible for this perfor-
computer networks have started to grad- service), current standards can no longer mance gain: The 'diversity gain' stabi-
ually replace regular wired Internet meet the requirements. lizes the link, as the additional antennas
access in homes, offices, and public can be used to compensate for a weak
areas, such as train stations and airports. Multiantenna (MIMO) systems are a connection on any of the other antennas.
However, as only limited bandwidth is means to increase the data rate, link reli- The 'array gain' allows the receiver to
allocated to be used by such systems, ability, and range of all kinds of wireless pick up more of the energy that was radi-
data rates need to be shared. As the networks and mobile communication ated from the transmitter, effectively
number of users grows and applications systems, without increasing bandwidth increasing the range of the system. The
The 16th European Conference on Artificial Intelligence (ECAI 2004) was held in
Valencia, Spain, from 22-27 August 2004, hosted by the Universidad Politécnica
de Valencia and the Grupo de Investigación de Tecnología Informática –Inteligencia
Artificial (GTI-IA). ECAI provided a public forum for researchers from academy
and industry and public organizations, offering them an invaluable occasion to
meet and exchange ideas, thus contributing to making real the upcoming
'Information Society'.
The conference included, from August Christian Freksa (University of Bremen, larly from Asia (53), North America (50)
22 to 24, 2004, two days and a half of Germany), Carole Goble (University of and Oceania (21). We can also notice the
workshops (27 workshops) and tutorials Manchester, UK), and Seppo Laukkanen emergence of high quality AI research in
(11 tutorials) on most advanced topics, (SenseTrix, Finland). South America (16) and Africa (8), and
as well as the second starting AI this is an indicator of the increasing
Researchers Symposium (STAIRS), Among 653 submissions received from interest in AI worldwide.
followed by the three-day technical 45 countries, 168 papers were accepted
program of ECAI. Following an estab- for oral presentation at ECAI 2004, 13 After the late registration deadline the
lished tradition, the sub-conference on for oral presentation at PAIS 2004, and, main conference (ECAI’04) has 694
prestigious applications of intelligent for the first time, 87 were accepted as registrations, STAIRS has 53 registra-
systems (PAIS) ran in parallel with posters. tions, and 436 people have registered for
ECAI. The technical programme the 27 workshops.
included refereed paper presentations, Although the majority of the submitted
More information:
invited talks by prestigious speakers and papers were from Europe (505), there https://www.dsic.upv.es/ecai2004/
poster sessions. The invited speakers has been a significant number of papers
were Glorianna Davenport (MIT, USA), from the rest of the world, and particu-
The results of the fifth campaign of the Cross-Language Evaluation Forum were
presented at a two-and-a-half day workshop held in Bath, UK, 15-17 September,
immediately following the eighth European Conference on Digital Libraries. The
workshop was attended by nearly 100 researchers and system developers.
The main objectives of the Cross- different aspects of mono- and cross- documents in a collection containing
Language Evaluation Forum (CLEF) language system performance. documents in eight languages, we felt
are to stimulate the development of that we had achieved an important goal.
mono- and multilingual information One of the principal objectives when We had shown that fully multilingual
retrieval systems for European CLEF began was to encourage devel- retrieval could be (almost) as effective
languages and to contribute to the opers to build multilingual retrieval as bilingual (L1 -> L2) retrieval and that
building of a research community in the systems capable of searching over systems are able to adapt and reengineer
multidisciplinary area of multilingual collections in a number of languages. rapidly and effectively to process new
information access. These objectives are The multilingual track was thus the languages as the need arises. For this
realised through the organisation of main track for several years and was reason, in CLEF 2004 we decided to
annual evaluation campaigns and work- made progressively more difficult. With reduce the multilingual document
shops. Each campaign offers a series of CLEF 2003, where the track included a retrieval activity to leave more space for
evaluation tracks designed to test task which involved finding relevant
other types of cross-language informa- Institute of Standards and Technology NTCIR evaluation initiative for Asian
tion retrieval experiments. (NIST), USA. languages, and on industrial experience
in building cross-language applications.
Six tracks were offered to evaluate the The response from the information The final session consisted of a panel in
performance of systems for: retrieval community was very encour- which panellists attempted to analyse
• mono-, bi- and multilingual document aging. Participation in this year’s the current organisation of the CLEF
retrieval on news collections (Ad-hoc) campaign was considerably up with campaigns in depth, considering
• mono- and cross-language domain- respect to the previous year with 55 whether we are working on the right
specific retrieval (GIRT) groups submitting results for one or problems, choosing our investments
• interactive cross-language retrieval more of the different tracks: 36 from wisely, and giving sufficient attention to
(iCLEF) Europe, 13 from N.America; 4 from the user perspective. Tracks taken into
• multiple language question answering Asia and one mixed European/Asian consideration for the CLEF 2005
(QA@CLEF) group. As in previous years, participants campaign include multilingual Web
• cross-language retrieval on image consisted of a nice mix of new-comers retrieval and a cross-language
collections (ImageCLEF) and veteran groups. The success of the Geographic Information Retrieval track.
• cross-language spoken document question answering and image retrieval
retrieval (CL-SDR). tracks had a big impact on participation The presentations given at the CLEF
in CLEF 2004, not just with respect to Workshops and detailed reports on the
CLEF 2004 thus marked a breaking the numbers but also regarding the skills experiments of CLEF 2004 and previous
point with respect to previous
campaigns. The focus was no longer
concentrated on multilingual document
retrieval but was diversified to include
different kinds of text retrieval across
languages (exact answers in the ques-
tion-answering track) and retrieval on
different kinds of media (not just plain
text but collections containing image
and speech as well). In addition,
increasing attention was given to issues
that regard system usability and user
satisfaction with tasks to measure the
effectiveness of interactive systems or
system components being included in
both the QA and the ImageCLEF tracks
with the collaboration of the coordina-
tors of iCLEF (the interactive track).
CALL FOR PAPERS • Enterprise Modelling Methods and CALL FOR PAPERS
Tools
CAiSE 2005 — The 17th • Service Oriented and Mobile Mirage 2005
Conference on Advanced Computing
• E-government Strategies and Applica- Rocquencourt (near Versailles),
Information Systems tions France, 1-2 March 2005
Engineering • Knowledge, Information and Data
Quality MIRAGE 2005 is an international
Porto, Portugal, 13-17 June 2005 • Web Content Management and conference with focus on Computer
Distribution Vision/Computer Graphics collaboration
Since the late 1980's, the CAiSE confer- • Workflow Systems techniques involving image analysis/
ences provide a forum for the presenta- • Knowledge Management synthesis approaches.
tion and exchange of research results and • Model and Software Reusability
practical experiences within the field of • Data Warehousing & OLAP In the domain of Computer Vision, this
Information Systems Engineering. • Metadata and Ontologies analysis by synthesis collaboration may
CAiSE’05 aims at bringing together • Support for Co-operative Work take the form of model-based approaches
researchers, users, and practitioners in • Innovative Database Technologies for which the analysis, recognition or
the field of information systems engi- • Distributed, Mobile, and Open understanding process does not rely on
neering. The conference programme will Architectures the data only but is linked to the opti-
feature paper presentations, workshops, • Agent-Oriented Technologies mization of a parametric model intro-
tutorials, and interactive panel sessions. • Languages and Protocols for IS duced A PRIORI. In the field of
• Component-Ware and IS Computer Graphics, this thematic refers
Conference Theme • IS Reengineering to image-based approaches which found
The Internet has been changing the • IS Usability and Interfaces to IS the modeling, animation, lighting or
society and the economy, and the way • Simulation in IS Development rendering process on the analysis of real
institutions and businesses operate has • Semantic Web. pre-existing data in order to gain in
evolved rapidly. All citizens are now productivity or in realism. In both cases,
expected to interact directly with tech- Case studies and experience reports this extraction and synthesis collabora-
nology based systems without direct related to the above topics are also tion can be iterative and conducted in a
human intermediation, using different welcome loop: the synthesized information laying
interface appliances. Moreover, many the foundations for a better characteriza-
information systems interact with each Important Dates tion that will in turn enable the synthesis
other with limited human supervision. • Paper Submission: 30 November 2004 step. Augmented Reality is a direct
The Internet and its infrastructure play a • Notification of Acceptance: 12 applition of this collaboration .
critical role in this evolution, and it is of February 2005
the utmost importance that semantic • Final Paper Submission: 12 March Authors are encouraged to submit papers
aspects are taken into consideration on a 2005 on theoretical, computational, experi-
sound basis. Improved communication mental or industrial aspects of model-
and understanding between people, the The conference is organized by FEUP - based image analysis and image-based
final objective of advanced information Faculdade de Engenharia da model synthesis.
systems, requires improved communica- Universidade do Porto (www.fe.up.pt)
tion and understanding between systems and DSIC-UPV - Departament de Topics
and its developers. Advanced informa- Sistemes Informàtics i Computació, Topics include (but are not limited to) :
tion systems engineering therefore needs Universitat Politècnica de València • model-based imaging and analysis
to keep improving its methods and tech- (www.dsic.upv.es) and sponsored by • iImage-based modeling and 3D recon-
nologies in order to support the proper ERCIM. struction
development of the economy and of our • data driven animation
More Information
society. http://www.fe.up.pt/caise2005/
• image and video-based lighting and
rendering
Relevant Topics • model-based vision approaches
In addition to the special theme, topics • model-based indexing and database
relevant for submissions to CAiSE 2005 retrieval
include, but are not limited to, the • model-based object tracking in image
following: sequences
• model-based image and shape analysis
• Methodologies, Models, and Tools for • model-based video compression tech-
IS Development niques.
• Requirements Engineering for IS
• Model Driven Architectures With applications in the field of :
• human/Computer interfaces
• video-games and entertainment CALL FOR PAPERS and services for the users, professionals,
industry and researchers in the field. Attendees
• mMedia production from film, broad- HCI International 2005 will be able to examine state-of-the-art
cast and games HCI technology and interact with manu-
• post-production, computer animation, Las Vegas, Nevada, USA, facturing representatives, vendors,
virtual special effects 22-27 July 2005 publishers, and potential employers.
• realistic 3D simulation , virtual proto- Exhibitors who commit to exhibit by
typing The 11th International Conference on November 20, 2004, apart from the fact
• multimedia applications, multimedia Human-Computer Interaction is jointly that they will have a better opportunity in
databases classification held under one management and one booth location selection and a very attrac-
• virtual and augmented reality registration with the Symposium on tive pricing, they will be listed in the
• medical and biomedical applications. Human Interface (Japan) 2005, the 6th Advance Program, Final Program, and
International Conference on Engineering the Conference Proceedings, which will
Important Dates Psychology & Cognitive Ergonomics, be distributed worldwide. Exhibitors who
• 15 November 15, 2004: Full paper the 3rd International Conference on commit to exhibit by May 1, 2005 will be
submission Universal Access in Human-Computer listed in the Final Program.
• 20 December 2004: Notification of Interaction, the 1st International
acceptance Conference on Virtual Reality, the 1st Thematic Areas
• 15 January 2005: Camera-ready International Conference on Usability • Ergonomics and Health Aspects of
papers due, early registration deadline and Internationalization, the 1st Work with Computers;
• 1 February 2005: Registration dead- International Conference on Online Program Chair: Pascale Carayon
line for authors of accepted papers Communities and Social Computing, • Human Interface and the Management
• 15 February 2005: Late registration and the 1st International Conference on of Information;
deadline Augmented Cognition. Program Chair: Michael J. Smith
• 1-2 March 2005. • Human-Computer Interaction;
HCI International 2005 and the affiliated Program Chair: Julie Jacko
More information:
http://acivs.org/mirage2005/
conferences will be held at the Caesars • Engineering Psychology and Cognitive
Palace Hotel, one of the most prestigious Ergonomics;
resorts in the world, under the auspices Program Chair: Don Harris
of 9 distinguished international boards of • Universal Access in Human-Computer
CALL FOR PARTICIPATION 202 members from 31 countries! Interaction;
Program Chair: Constantine Stephanidis
EGC2005 — The conference objective is to provide an • Virtual Reality;
European Grid Conference 2005 international forum for the dissemination Program Chairs: Kay Stanney &
and exchange of scientific information Michael Zyda
Amsterdam, The Netherlands, on theoretical, generic, and applied areas • Usability and Internationalization;
14-16 February 2005 of HCI, usability, internationalization, Program Chair: Nuray Aykin
virtual reality, universal access and • Online Communities and Social
The European Grid Conference cognitive ergonomics. This will be Computing;
EGC2005 is the premier event on Grid accomplished through the following six Program Chair: Jennifer Preece
Computing in Europe. It will focus on all modes of communication: plenary • Augmented Cognition;
aspects of Grid computing in Europe and presentation, parallel sessions, demon- Program Chair: Dylan D. Schmorrow.
a such will bring together participants stration and poster sessions, tutorials,
from research and industry. EGC 2005 is exhibitions and meetings of special Important Dates
a follow-up of the AcrossGrids interest groups. The six-day conference • Paper Presentations - Deadline for
Conferences held in Santiago de will start with three days of tutorials. The Abstract Receipt: 1 October 2004
Compostela (2003) and in Nicosia tutorials will begin on Friday, 22 July • Posters/demonstrations - Deadline for
(2004). 2005. A total of 18 tutorials will be Abstract Receipt: 1 April 2005
offered (both half-day and full-day) at • Special Interest Groups - Deadline for
EGC2005 will have three main tracks: introductory, intermediate, and Abstract Receipt: 20 October 2004
• a scientific track; advanced levels covering the entire spec- • Tutorials - Deadline for Abstract
• an industrial track; trum of the 9 tracks and 7 conferences. Receipt: 20 October 2004
• a special events track. • Early Registration: 20 January 2005
HCI International conference gathers • Mid Registration: 21 January - 20 May
Together they aim to cover the complete over a thousand scientists, academics and 2005
field of Grid developments in Europe. professionals from research institutions, • Late Registration: After 20 May 2005
universities, and companies from all over
More information: More information:
http://genias.biz/egc2005
the world. The HCI International 2005 http://www.hci-international.org/
exhibits will provide state-of-the-art HCI,
VR, psychology and computer products
ECDL 2005 — European Virtual Reality 2005 INTERACT 2005 — Tenth IFIP
Conference on Digital Libraries TC13 International Conference
Bonn, Germany, 12-16 March 2005
Vienna, Austria,
on Human-Computer Interaction
18-23 September 2005 IEEE VR 2005 is the premier interna- Rome, Italy,
tional conference and exhibition in 12-16 September 2005
ECDL 2005 is the 9th conference in the virtual reality, and provides a truly
series of European Digital Library unique opportunity to interact with The Interact '05 conference will highlight
conferences. ECDL has become the leading experts in VR and closely related to both the academic and industrial world
major European conference on digital fields such as augmented reality, mixed the importance of the Human-Computer
libraries, and associated technical, prac- reality and 3D interaction. Test your own Interaction area and its most recent
tical, and social issues, bringing together work and educate yourself through expo- breakthroughs on current applications.
researchers, developers, content sure to the research of your peers from
providers and users in the field. ECDL around the world. And of course, there Topics:
2005 is jointly organized by the Vienna will be time to renew friendships, make Suggested topics include but are not
University of Technology (VUT), the new ones and experience the European limited to:
Austrian National Library (ÖNB), and cuisine and culture in Bonn, the former • Multi-modal Interfaces
the Austrian Computer Society (OCG). German capital. • Context-dependent Systems
The conference will take place in the • End User Development
Vienna Technical University. IEEE VR is an excellent opportunity to • Intelligent Environments
present: • Tangible Interfaces
Important Dates: • leading edge research contributions • Novel User Interfaces
• workshop proposal deadline: (papers) • Usability Evaluation
28 January 2005 • innovative applications and experi- • Location-aware Interaction
• workshop acceptance notification: ence reports (application sketches) • Model-based Design of Interactive
27 February 2005 • communicate the state of the art Systems
• paper/tutorial/panel submission dead- (survey reports) • Multi-user Interaction
line: 1 March 2005 • present and discuss solutions (panels) • Safety Issues and Human error
• acceptance notifications: 15 May 2005 • introduce to your VR research (lab • Cognitive Models
• final version of papers: 3June 2005. presentations). • Visualisation Techniques
• Web Design and Usability
Topics: The main conference starts with tutorials • Mobile and Ubiquitous Computing
Topics of contributions include (but are and workshops on March 12 and 13. • 3D and Virtual Reality Interfaces
not limited to): They will introduce to specific topics of • Adaptive Interfaces
• concepts of digital libraries and digital the field and give a compact view of • Web Accessibility
documents ongoing research activities. • Usability & Accessibility.
• system architectures, integration and
interoperability The industrial exhibition and the Important Dates:
• information organization, search and research lab demonstrations will provide Full papers, tutorials, workshops,
usage practical insights into solutions and doctoral consortium:
• user studies and system evaluation prototypes. • Submission deadline: 31 January 2005
• digital preservation • Notification of acceptance: 15 March
• digital library applications. IEEE VR2005 is cohosted by: 2005
• caesar: center for advanced European • Final version due: 30 April 2005.
More information:
http://www.ecdl2005.org
studies and research
• University of Applied Sciences Bonn- Short papers, special interest groups,
Rhein-Sieg panels, posters, demos:
• Fraunhofer Institute for Applied • Submission deadline: 30 March 2005
Information Technology • Notification of acceptance: 10 May 2005
• Rheinische Friedrich-Wilhelms- • Final version due: 30 May 2005.
Universitaet Bonn.
The conference is sponsored by ERCIM.
The conference is sponsored by ERCIM
More information:
More information: http://www.interact2005.org/
http://www.vr2005.org
ECIR'05 — 27th European Workshop on Construction The workshop will focus on:
Conference on Information and Analysis of Safe, Secure • Research trends in smart devices
• Web services
Retrieval and Interoperable Smart • Virtual machine technology
Santiago de Compostela, Spain, Devices • Security
21-23 March 2005 • Proof-Carrying Code
Nice, France, 7-11 March 2005 • Validation and formal methods
The annual European Conference on • Embedded devices.
Information Retrieval Research, orga- The CASSIS International Workshop
nized by the Information Retrieval will be held in the week of March 7th to The workshop is organized by the
Specialist Group of the British Computer March 11th 2005 in Nice and will consist INRIA's Everest team.
Society (BCS-IRSG), is the main of keynote and invited presentations by
More information:
European forum for the presentation of leading academics and industrials in the http://www-sop.inria.fr/everest/events/
new research results in the field of infor- field of security and analysis of smart cassis05/
mation retrieval. devices.
ISTI-CNR — Paolo Cignoni is the SpaRCIM — The First Spanish Conference on Informatics (CEDI 2005) is
first recipient of the Eurographics 'Young to take place between 13-16 September, 2005. While national conferences are tradi-
Researcher' award in recognition of his tionally organized in several specific topics, in this case, all are to take place simulta-
outstanding contributions to the develop- neously and in the same location. The main goal is to join the Spanish informatics
ment and use of advanced multiresolution research community in order to demonstrate scientific advances, discuss specific
techniques in computer graphics and problems and improve the visibility of this area in Spain, while emphasizing its role in
scientific visual- the age of the information society.
ization.
CEDI will be organized in a 'federated conference' format, joining several more
The scope of specific symposia in areas including Artificial Intelligence, Software Engineering and
Cignoni's inter- Databases, Programming Languages, Parallelism and Compute Architecture,
ests ranges from Computer Graphics, Concurrency, Soft-Computing, Pattern Recognition, Natural
scientific visual- Language Processing, Ubiquitous Computing and Human-Computer Interaction.
ization to LOD
techniques. He In addition to the scientific activities of each symposium, the conference will organize
has published a number of invited talks, round tables, social events and a gala dinner, at which
more than 60 several awards will be announced, recognizing the efforts of some colleagues in
research papers in international refereed promoting the growth of the area in Spain.
journals and conferences, including six
papers at Eurographics, and three papers More than 1000 researchers are expected to attend the conference. The event will be
at Siggraph'04. In addition, his work has funded by the Spanish Ministry of Education and Science, and is sponsored by
found important applications in 3D scan- SpaRCIM.
ning and cultural heritage.
The conference is organized by the Department of Computer Science at the
Among his more recent results there is a University of Granada. It will be chaired by Alberto Prieto from the University of
novel approach to variable resolution Granada, with Juan José Moreno Navarro (UPM and SpaRCIM director) as Program
rendering based on a 'GPU-friendly' data Chair. Further details can be found at http://cedi2005.ugr.es/
structure that is one or two orders of
magnitude faster than previous solutions
and has been proposed for the efficient
rendering of planet-sized digital terrains CWI — CWI has started a new line of research on fundamental and practice-
and huge 3D models. oriented cryptology. The institute is already known for the Number Field Sieve project
for factoring large integers, which attracted world-wide attention with the factorization
Paolo Cignoni is one of the founding of the RSA-512 internet security code. The new Cryptology and Information Security
members of the Visual Computing Lab at group, led by prof. Ronald Cramer, includes this project. Additionally research is initi-
ISTI-CNR and a lecturer at the Computer ated in computational number theory with relevance for cryptology, public key cryptog-
Science Dept. of the University of Pisa. raphy, information theoretically secure cryptography, quantum cryptography, formal
security analysis, and applied information security. One of the group's major focal
points is secure computation. This area deals with two or more parties who wish to
achieve a joint task securely even though they are mutually distrustful and wish to keep
INRIA — Gilles Kahn, Chairman sensitive, private information secret from each other. This is sometimes called multi-
of INRIA and INRIA's representative on lateral security, as opposed to unilateral security in the case of secure communications.
ERCIM's board of directors was Practical applications include profile matching, joint database comparison, electronic
appointed member at the French National voting or auctions, and threshold cryptography. More information can be found on
Strategic Advisory Board on Information http://www.cwi.nl/pna5
Technologies on 27 September 2004. The
Strategic Advisory Board on Information
Technologies (CSTI — Conseil
Stratégique des Technologies de FNR — The National Research Fund of Luxembourg launched the second
l'Information) is chaired by the French call of its multiannual research programme 'Security and efficiency of new practices
Prime Minister and composed of leading in e-commerce for all socio-economic actors' (SECOM) on April 1st 2004. It has been
entrepreneurs from industry and R&D, adopted by the Council of Government in the year 2000 and covers a total budget of
selected for their own competence. CSTI EUR 1.2 million over the period 2005 to 2007. Five project proposals were submitted by
is responsible for presenting recommen- the deadline of July 15th 2004 and are now being evaluated by independent experts. The
dations to Government related to strategic selected projects will start early 2005.
orientations in innovation and research
and development, concerning information
technologies.
Council for the Central Laboratory of the Research Norwegian University of Science and Technology
Councils, Rutherford Appleton Laboratory Faculty of Information Technology, Mathematics and
Chilton, Didcot, Oxfordshire OX11 0QX, United Kingdom Electrical Engineering, N 7491 Trondheim, Norway
Tel: +44 1235 82 1900, Fax: +44 1235 44 5385 Tel: +47 73 59 80 35, Fax: +47 73 59 36 28
http://www.cclrc.ac.uk/ http://www.ntnu.no/
Consiglio Nazionale delle Ricerche, ISTI-CNR Spanish Research Consortium for Informatics
Area della Ricerca CNR di Pisa, and Mathematics
Via G. Moruzzi 1, 56124 Pisa, Italy Esperanza Marcos, Rey Juan Carlos University,
Tel: +39 050 315 2878, Fax: +39 050 315 2810 C/ Tulipan s/n, 28933-Móstoles, Madrid, Spain,
http://www.isti.cnr.it/ Tel: +34 91 664 74 91, Fax: 34 91 664 74 90
Centrum voor Wiskunde en Informatica Swiss Association for Research in Information Technology
Kruislaan 413, NL-1098 SJ Amsterdam, c/o Prof. Dr Alfred Strohmeier, EPFL-IC-LGL,
The Netherlands CH-1015 Lausanne, Switzerland
Tel: +31 20 592 9333, Fax: +31 20 592 4199 Tel +41 21 693 4231, Fax +41 21 693 5079
http://www.cwi.nl/ http://www.sarit.ch/
✄
Order Form I wish to subscribe to the
If you wish to subscribe to ERCIM News ❏ printed edtion ❏ online edition (email required)
free of charge
or if you know of a colleague who would like to Name:
receive regular copies of
ERCIM News, please fill in this form and we Organisation/Company:
will add you/them to the mailing list.
Address:
Send, fax or email this form to:
ERCIM NEWS
2004 route des Lucioles
BP 93 Post Code:
F-06902 Sophia Antipolis Cedex
Fax: +33 4 9238 5011 City:
E-mail: office@ercim.org
Country
Data from this form will be held on a computer database.
By giving your email address, you allow ERCIM to send you email E-mail:
You can also subscribe to ERCIM News and order back copies by filling out the form at the ERCIM website at
http://www.ercim.org/publication/Ercim_News/