You are on page 1of 10

International Journal of Advance Foundation and Research in Computer (IJAFRC)

Volume 1, Issue 6, June 21!" I##$ 2%!& ' !&(%


1) * + 21!, IJAFRC All Ri,hts Reserved ---"i.afrc"or,
An Anal/sis 0f Cloud Relia1ilit/Approaches 2ased on Cloud
Components And Relia1ilit/ 3echni4ues
Abishi Chowdhury
1
, Priyanka Tripathi
2
National Institute of Technical Teachers Training and Research !hopal India
12
abishi"chowdhury#g$ail"co$
1
ptripathi#nitttrbpl"ac"in
2

A 2 # 3 R A C 3
Cloud computin, is po-erin, the overall 1usiness and or,ani5ational ,ro-th 1/ providin, the
three1asic services li6e #oft-are as a #ervice (#aa#), 7latform as a #ervice (7aa#) and
Infrastructure as a #ervice (Iaa#)" As the -orld-ide cloud users demand multiple services from
cloud at a time, so it is the most important concern for cloud service providers to concentrate on
the relia1ilit/ of the s/stem" 3he relia1ilit/ of an/ s/stem can 1e determined 1/ the num1er of
failures occurred in the cloud computin, environment vs" the total num1er of tas6s done 1/ the
cloud" 3he relia1ilit/ of a s/stem depends on the relia1ilit/ of each and ever/ component of the
s/stem -ith -hich the s/stem is composed of" In this paper, an attempt has 1een made to anal/5e
different cloud relia1ilit/ techni4ues, different components for relia1ilit/ measurement and the
methodolo,/ for measurin, relia1ilit/" Further, 1ased on these parameters -e have prepared a
comparison ta1le to compare these techni4ues"
Inde8 3erms9 Virtual :achine; Relia1ilit/; Cloud :ana,er; Fault 3olerance; Fault mana,er; Fault
3olerance middle-are

I" I$3R0<=C3I0$
Cloud co$puting pro%ides on de$and ser%ices to its users and the users can de$and any kind of ser%ices
in the for$ of &oftware Platfor$ Infrastructure and so on at anyti$e fro$ anywhere '1('2(" Cloud
reser%es its abstract nature while pro%iding these ser%ices to the cloud users" Cloud co$prises different
ser%ers" A )atacenter can be a collection of thousands of the ser%ers" *sers re+uest for the cloud
infrastructure and use the ser%ers for doing their tasks" The cloud pro%ides infrastructure in the for$ of
%irtual $achines ,-.s/" Cloud can also pro%ide a whole %irtual infrastructure for pro%iding ser%ices to
the cloud users" And for doing this cloud uses different types of approaches"
Any type syste$ can go under failure" 0ailure in the cloud en%iron$ent is not an e1ceptional case because
fault is a nature of the technology" 0ailure can occur at any ti$e" 0ailure affects different aspect of the
syste$" .ost i$portantly it affects the %ast worldwide business of the cloud co$puting" &o$eti$es a
s$all failure can gi%e a great loss to the cloud ser%ice pro%ider" 0ailure can affect the re%enue and the
long ter$ i$age of the cloud ser%ice pro%iders"
0ailures can be hardware failures and the software failures" !oth re+uire different strategy for getting a
solution" 2ardware failure can be the failure like failure of the $e$ory failure of the disk etc" The
software failure can be like the failure application failure 31ecution ti$e failure Ti$eout failure etc"
Reliability of cloud resources does not depend on the reliability of the indi%idual resources" !ut it
depends on the reliability of their collecti%e working" 4hile calculating the reliability of cloud co$puting
it should be kept in $ind that whether the co$ponents are working in parallel or not"
This paper presents a study of cloud reliability and the techni+ues which are proposed for $easuring and
i$pro%ing the cloud reliability" 0irst the National Institute of &tandards and Technology ,NI&T/ standards
International Journal of Advance Foundation and Research in Computer (IJAFRC)
Volume 1, Issue 6, June 21!" I##$ 2%!& ' !&(%
1& * + 21!, IJAFRC All Ri,hts Reserved ---"i.afrc"or,
about the reliability of cloud co$puting are discussed" And then different techni+ues about $easuring
the cloud reliability and the techni+ues to o%erco$e the faults in the cloud en%iron$ent are discussed"
II" R>?IA2I?I3@ C0$C>73

A" $I#3 #tandards
As stated in National Institute of &tandards and Technology NI&T '5( broadly reliability is the function of
four $ain co$ponents of cloud co$puting"
6 The software and hardware offered by the cloud ser%ice pro%iders
6 The personnel resources pro%ided
6 Connecti%ity
6 The consu$ers personnel
It is %ery difficult task to $easure the reliability of cloud co$puting en%iron$ent" The $ain reasons for
this first as there are a nu$ber of co$ponents in a cloud en%iron$ent the indi%idual reliability of these
co$ponents is different fro$ the reliability of these co$ponents taken together" And the second is that it
is highly dyna$ic and depends on the en%iron$ent" Now first we ha%e to consider all the possible
conditions of failures in cloud en%iron$ent then only a reliability $odel can be considered"
2" <IFF>R>$3 R>?IA2I?I3@ A77R0ACA>#
1" Adaptive Fault 3olerance (AF3)
It is an adapti%e fault tolerance techni+ue '7( in real ti$e cloud co$puting en%iron$ent ,A0TRC/" It
tolerates the faults on the basis of reliability of each %irtual $achine" !ased on the reliability a %irtual
$achine is selected" There are two types of nodes the %irtual $achine which are running on the cloud
and the ad8udication node" 9n the running %irtual $achine we ha%e the real ti$e application and the test
for acceptance of logical %alidity" At ad8udication node there is the decision syste$ reliability assessor
and ti$e checker" In brief this techni+ue uses
6 Acceptance test: This is for checking the results of the real ti$e algorith$s"
6 Ti$e checker: It checks the ti$e of the results produced by each $odule"
6 Reliability assessor: It assesses the reliability of e%ery %irtual $achine"
6 )ecision $echanis$: It is used for $aking decision about %irtual $achine"
6 Reco%ery cache: It is used for the checkpoints"
2" Cloud #ervice Relia1ilit/ :odellin, and Anal/sis
In this paper the authors had presented ';( an in%enti%e reliability $odel for cloud co$puting which
deals with se%eral types of failures that affect the success and failure of cloud ser%ices" 0irst a cloud
co$puting syste$ in the -<rA)& ,-irtual <rid Application )e%elop$ent &oftware/ is proposed" In this
syste$ there is a C.& ,Cloud .anage$ent &yste$/ which is co$posed of a set of ser%ers that ser%es four
different responsibilities" &uch as
6 .anaging a re+uest +ueue that contains the re+uests of different cloud users
6 .anaging co$puting resources such as PCs etc"
6 .anaging data resources such as )atabases etc"
6 &cheduling re+uests and assigning these to different co$puting and data resources"
4hene%er any re+uest co$es it passes through the C.& then C.& pro%ides the resources to the$" A
nu$ber of failures were analy=ed such as Co$puting resource $issing )ata resource $issing 9%erflow
International Journal of Advance Foundation and Research in Computer (IJAFRC)
Volume 1, Issue 6, June 21!" I##$ 2%!& ' !&(%
1B * + 21!, IJAFRC All Ri,hts Reserved ---"i.afrc"or,
failure ti$eout failure )atabase failure Network failure &oftware failure and 2ardware failure
'>('?('@(" Classification of reliability stages is gi%en in figure1"

Fi,ure 1" Relia1ilit/ #ta,es
The solution proposed for the Re+uest &tage Reliability is the .arko% .odel and for 31ecution &tage
Reliability is a graph based $odel" This $odel is further enhanced with the co$bination of the graph
based $odel and the !ayesian network $odel" The o%erall reliability is gi%en by the $ultiplication of the
re+uest stage reliability and the e1ecution stage reliability"
%" Fault'3olerant and Relia1le Computation in Cloud Computin,
In this paper the authors ha%e e1plored the security aspect of scientific co$putation in cloud co$puting"
The proper cloud selection strategy and protection against faulty and $ischie%ous cloud was in%estigated
'A(" They ha%e considered the scientific co$putation in large $atri1 $ultiplication" At first they assu$ed
that there are se%eral clouds and each of which contains se%eral ser%ers" These ser%ers are trusted
partially based on the e1perience of the indi%idual client and the client knows the reliability and cost of
each of the cloud" )ifferent cost for different cloud" The work is di%ided in the $ultiplication of the rows
and colu$ns on the different clouds" The cost calculation of different clouds is gi%en" Now the proble$ is
suppose a client dispatches li rows of $atri1 A for being $ultiplied with $atri1 ! in the cloud i" Then the
o%erall cost will be:
C B C

L
=1
l


The reliability of the dispatched task will be:
RB , R
L
=1

l

/C l
4here l B no of rows in $atri1 A and l

= l
L
=1

The $ain ob8ecti%e is to $ini$i=e the o%erall cost C sub8ect to R DB Rs where Rs is a $ini$u$ reliability
re+uire$ent which is pre%iously specified" Now the o%erall reliability of this task can be e1pressed as the
$ini$u$ %alue of the reliability of all the clouds in%ol%ed in this calculation i"e"
R min
=1,2,L
{R

]
Then we can si$ply discard all the clouds with reliability less than the $ini$u$ re+uired reliability Rs"
Now fro$ the re$aining set of the clouds we can choose the cloud with reliability %alue greater than Rs
and with the lowest cost Ci first" Then we can choose other clouds with higher reliability as per necessity"
International Journal of Advance Foundation and Research in Computer (IJAFRC)
Volume 1, Issue 6, June 21!" I##$ 2%!& ' !&(%
11 * + 21!, IJAFRC All Ri,hts Reserved ---"i.afrc"or,
!" Fault 3olerance and Resilience
In this paper '1E( the concept of fault errors and failure can be e1pressed by applying the following
chain:
0ault 3rror 0ailure
The failure beha%ior of the ser%ers that are contained in the data center can be obtained by the study
about the ser%er failures and the hardware failures" It is necessary to apply fault tolerance syste$ to
enhance the reliability of hard disks in order to considerably cut down the nu$ber of failures" According
to the study of the syste$ failed $achines are replaced" The study of the failure beha%ior of networks
should also be done as se%eral network co$ponents are associated for constructing the data center"
!ased on this study it is obser%ed that the reliability of co$prehensi%e data center network is al$ost
AA"AAF" 0ault tolerance is the capability of the syste$ to achie%e its function in spite of the presence of
failures" The classification of faults are done into two categories as shown in the below figure 2:

Fi,ure 29 Classification of Faults
0irst Crash faults which block the se%eral syste$ co$ponents fro$ functioning or to re$ain idle at the
ti$e of failures for e1a$ple hard disk crash power outage etc"
&econd is 2/5antine faults that cause the syste$ co$ponents to beha%e incorrectly at the ti$e of failure"
As a result the syste$ shows erratic beha%ior"
The $ost popular $ethods to resol%e these abo%e two types of faults are described below:
Chec6in, and monitorin,9 In this $ethod the syste$ is being obser%ed continuously during its runti$e
in order to 8ustify the correctness of the syste$ specification"
Chec6point and restart9 In this $ethod the state of the syste$ is grabbed and stored so that when the
syste$ goes through a failure its correct state is restored using the checkpoint infor$ation"
Replication9 And in this $ethod the essential syste$ co$ponents are replicated or i$itated in such a
way so that a copy of this syste$ co$ponents is a%ailable during a failure"
(" Fault 3olerance :iddle-are
The Gow latency 0ault Tolerance ,GG0T/ $iddleware '11( is a ser%ice that contributes fault tolerance
reliable ser%ices for distributed applications within data centers that co$prises of se%eral ser%ers
storage and networks" !y using leaderCfollower approach this GG0T $iddleware i$itates the application
process in order to secure the application fro$ se%eral faults particularly the Crash fault and Ti$ing
fault" )ue to crash fault a process or processor does not yield any further result and it does not yield any
result within a specific ti$e constraint due to the ti$ing fault" !ut the !y=antine fault is not handled by
this $iddleware" Two types of leaderCfollower replications are supported by this GG0T $iddleware"
These are as follows:
#emi active replication9 In this process the pri$ary replica orderly arranges the recei%ed
$essages and e1ecutes the operations and also pro%ides the ordering infor$ation to the backup
replicas for the nonHdeter$inistic operations"
International Journal of Advance Foundation and Research in Computer (IJAFRC)
Volume 1, Issue 6, June 21!" I##$ 2%!& ' !&(%
111 * + 21!, IJAFRC All Ri,hts Reserved ---"i.afrc"or,
#emi passive replication9 It perfor$s not only the sa$e of the abo%e but also in addition to this
the pri$ary replica always co$$unicates state update to the backup replicas" It uses lesser
processing power than se$i acti%e replication but if the pri$ary fails it ac+uires greater latency
for the purpose of reco%ery and reconfiguration"
The GG0T $iddleware co$prises of the following:
?o- ?atenc/ Fault 3olerance (??F3) :essa,in, 7rotocol9 It basically contributes two $ain
ser%ices for application $essagesI these are the following:
Relia1le deliver/9 In this all the $e$bers of a group recei%e each and e%ery $essage that
is $ulticast to this group on a network connection"
3otal orderin,9 The pri$ary replica in a group co$$unicates the ordering infor$ation
to all the backups in this group and all the $e$bers in a group hand o%er the $essages to
the application $aintaining the sa$e order"
?o- ?atenc/ Fault 3olerance (??F3) :em1ership 7rotocol9 This protocol confir$s that all the
$e$bers of a particular group $ust ha%e a consistent %iew about the $e$bership set and the
pri$ary replica of that group" It is $uch faster than a $ultiHround consensus protocol which is
$ainly necessary in the case when pri$ary replica fails" The pri$ary replica decides the inclusion
and e1clusion of the backup replicas to or fro$ the group on the basis of their ranks and
precedence"
The precedence of a group $e$ber is deter$ined by the order in which the $e$bers 8oin the
group and the rank of the pri$ary replica is 1 and for the backup replicas it will be 2 5 7J which
are assigned by the pri$ary replica based on their precedence"
?o- ?atenc/ Fault 3olerance (??F3) Virtual <etermini5er Frame-or69 The applications in
cloud co$puting en%iron$ent co$$only incorporate se%eral sources of nonHdeter$inis$"
Therefore to preser%e fir$ replica consistency it is %ital to $ask these sources of nonH
deter$inis$" It records the ordering infor$ation and the results of each non deter$inistic
operation acco$plished by the pri$ary replica and at the back up replicas it carries out the sa$e
ordering as the pri$ary"
6" A #/stem ?evel Approach
The purpose of this approach '12( is to o%erco$e the li$itations of current e1isting $ethodologies by
pro%iding fault tolerance properties as on de$and ser%ices" It contributes fle1ibility for the applications
to dyna$ically regulate its fault tolerance properties and the le%el of the re+uired a%ailability and
reliability o%erti$e" The cost of the resources can be reduced to a certain e1tent and the perfor$ance
le%el can be ad8usted according to the particular business needs" It allows the users to obtain an e1plicitly
fault tolerance support for its applications without ha%ing a co$prehensi%e knowledge about the syste$
le%el proceedings" !y adding a new dedicated ser%ice layer between the co$puting fra$ework and the
applications it is possible to pro%ide fault tolerance reliable support to each application abstracting the
co$plications of the ele$ental infrastructure" To pro$ote a wellHde%eloped support it is necessary for
the ser%ice layer to acco$$odate a range of reliability $echanis$s and also to construct a fault tolerance
solution which can be dispatched to different applications" And to acco$plish this fault tolerance
solution can be %iewed as a co$bination of a set of definite acti%ities" 0or e1a$ple each fault tolerance
$echanis$ such as fault detection replication of an application $asking and reco%ery etc" can be
obser%ed as a specific or distinct acti%ity that are co$bined together to build a fault tolerance solution"
Now this each indi%idual acti%ity can be acco$plished as a standHalone configurable $odule which
produces a consistent solution to a repetiti%e syste$ failure" .oreo%er each $odule is co$bined with a
set of $etadata which characteri=e its functional structural and operational properties" These $etadata
International Journal of Advance Foundation and Research in Computer (IJAFRC)
112 * + 21!, IJAFRC All Ri,hts Reserved
can be inspected during runti$e and co$pared with diffe
rele%ant acti%ities" This approach can be achie%ed by i$ple$enting each $odule separately as a web
ser%ice in the for$ of 4&)G '15(
designing a sche$e the 0ault Tolerance .anager ,0T./"
0T. is co$posed of a set of following co$ponents:
6 Replication :ana,er9 It incorporates techni+ues to $anage fir$ness in a replica group by updating
the state of the backup replicas and the pri$ary replica"
6 Fault <etectionC7rediction :ana,er9
and to send notification regarding this to 0T. kernel in order to in%oke ser%ices fro$ 0ault .asking
.anager and Reco%ery .anager" It also notifies Resource .an
the resource state of the cloud"
6 Fault :as6in, :ana,er9 This co$ponent in%ol%es a collection of algorith$s that $ask the
occurrence of failures and restrict the faults to $eet high a%ailabil
6 Recover/ :ana,er9 It incorporates all the $echanis$s which is used to resu$e er
a nor$al node"
6 :essa,in, :onitor9 It is used to con%ey necessary $essages a$ong different replicas of a replica
group and also for interHco$ponent c
6 ClientCAdmin Interface9 This is used to achie%e users re+uire$ents and act as an interface
0T. and the end users"
6 F3: Dernel9 It is the pi%otal co$puting co$ponent of 0T. which $anages the reliability $
present in the sche$e"
6 Resource :ana,er9 This co$ponent is used to efficiently allocate re+uired resources and to a%oid
under pro%isioning and o%er pro%isioning during failures"

)" Fault 3olerant Approaches in Cloud Infrastructure
In $ost of the recent approaches fault
custo$ers" !ut there are no collaborations between the$" Therefore so$eti$es this leads to a partial or
faulty solution" To o%erco$e this issue different fault tolerance policies in clo
in%estigated in this paper '17(" There are $ainly two types of policies i"e" in the first one fault tolerance
$echanis$s are handled solely by either the cloud ser%ice pro%ider or the custo$er and in the second
policy there is a collaborati%e $anage$ent between the custo
Fi,
In general there are three layers in a cloud platfor$ that is shown in the figure
-irtual $achines and Resources and
International Journal of Advance Foundation and Research in Computer (IJAFRC)
Volume 1, Issue 6, June 21!"
+ 21!, IJAFRC All Ri,hts Reserved
can be inspected during runti$e and co$pared with different users re+uire$ents in order to choose
rele%ant acti%ities" This approach can be achie%ed by i$ple$enting each $odule separately as a web
docu$ent" The feasibility of this proposed approach is obtained by
he$e the 0ault Tolerance .anager ,0T./"
0T. is co$posed of a set of following co$ponents:
It incorporates techni+ues to $anage fir$ness in a replica group by updating
plicas and the pri$ary replica"
t <etectionC7rediction :ana,er9 It is used to detect the faults pro$ptly after their occurrence
and to send notification regarding this to 0T. kernel in order to in%oke ser%ices fro$ 0ault .asking
.anager and Reco%ery .anager" It also notifies Resource .anager about the faulty replica to update

This co$ponent in%ol%es a collection of algorith$s that $ask the
occurrence of failures and restrict the faults to $eet high a%ailability de$ands of the cloud
It incorporates all the $echanis$s which is used to resu$e er
It is used to con%ey necessary $essages a$ong different replicas of a replica
co$ponent co$$unication"
This is used to achie%e users re+uire$ents and act as an interface
It is the pi%otal co$puting co$ponent of 0T. which $anages the reliability $
This co$ponent is used to efficiently allocate re+uired resources and to a%oid
under pro%isioning and o%er pro%isioning during failures"
Fault 3olerant Approaches in Cloud Infrastructure
In $ost of the recent approaches fault tolerance is entirely handled by the cloud ser%ice pro%iders or the
custo$ers" !ut there are no collaborations between the$" Therefore so$eti$es this leads to a partial or
faulty solution" To o%erco$e this issue different fault tolerance policies in cloud co$puting ha%e been
" There are $ainly two types of policies i"e" in the first one fault tolerance
$echanis$s are handled solely by either the cloud ser%ice pro%ider or the custo$er and in the second
laborati%e $anage$ent between the custo$ers and the ser%ice pro%iders"
Fi,ure %" Cloud computin, Architecture
In general there are three layers in a cloud platfor$ that is shown in the figure 5
-irtual $achines and Resources and each of these are associated with se%eral failures" That is why there
International Journal of Advance Foundation and Research in Computer (IJAFRC)
Volume 1, Issue 6, June 21!" I##$ 2%!& ' !&(%
---"i.afrc"or,
rent users re+uire$ents in order to choose
rele%ant acti%ities" This approach can be achie%ed by i$ple$enting each $odule separately as a web
" The feasibility of this proposed approach is obtained by
It incorporates techni+ues to $anage fir$ness in a replica group by updating
It is used to detect the faults pro$ptly after their occurrence
and to send notification regarding this to 0T. kernel in order to in%oke ser%ices fro$ 0ault .asking
ager about the faulty replica to update
This co$ponent in%ol%es a collection of algorith$s that $ask the
ity de$ands of the cloud users"
It incorporates all the $echanis$s which is used to resu$e erroneous nodes to
It is used to con%ey necessary $essages a$ong different replicas of a replica
This is used to achie%e users re+uire$ents and act as an interface between
It is the pi%otal co$puting co$ponent of 0T. which $anages the reliability $echanis$
This co$ponent is used to efficiently allocate re+uired resources and to a%oid
tolerance is entirely handled by the cloud ser%ice pro%iders or the
custo$ers" !ut there are no collaborations between the$" Therefore so$eti$es this leads to a partial or
ud co$puting ha%e been
" There are $ainly two types of policies i"e" in the first one fault tolerance
$echanis$s are handled solely by either the cloud ser%ice pro%ider or the custo$er and in the second
$ers and the ser%ice pro%iders"

5 these are Applications
each of these are associated with se%eral failures" That is why there
International Journal of Advance Foundation and Research in Computer (IJAFRC)
11% * + 21!, IJAFRC All Ri,hts Reserved
are $ainly three types of failures: Application failure -irtual $achine failure and 2ardware failure" And
for these failures there are so$e fault tolerance so
0irst fault tolerance $ethod concentrates in stateless applications like pro1y e"g" 2APro1y or .y&KG
Pro1y" The second is a stateHfull $ethod in this custo$er $ust i$ple$ent the functions for storing the
state of the ser%er so that on the ne1t start of t
fault sensors can be used" 0irst the faulty -. is deallocated fro$ the 8ob" &econd a new -. is allocated"
Third start the tasks that are running on the failed -." 0ourth restore the state of th
0ault tolerance syste$" The custo$er cannot see all these types of the fault" These can be $onitored
the cloud ser%ice pro%iders" This is done by a $onitoring syste$ co$posed of the sensors" These
techni+ues are used in '1;('1>("
&" A Virtuali5ation and Fault 3olerance Approach
0ault tolerance is pro%ided to cloud infrastructure by i$ple$enting the cloud $anager '1?( Goad
balancer 0ault 2andler and the )ecision $aker" A para$eter success rate is used specially for fault
tolerance" In this Lob is gi%en to the %irtual $achine which has the success rate $ore than so$e specific
%alue" In this way the chances of the fault decrease" The fault handler has the responsibility that when a
-. is found to be faulty its perfor$ance table $ust b
perfor$ance table the cloud infrastruct
International Journal of Advance Foundation and Research in Computer (IJAFRC)
Volume 1, Issue 6, June 21!"
+ 21!, IJAFRC All Ri,hts Reserved
are $ainly three types of failures: Application failure -irtual $achine failure and 2ardware failure" And
for these failures there are so$e fault tolerance solutions that are described in figure 7
Fi,ure !" Fault 3olerance
0irst fault tolerance $ethod concentrates in stateless applications like pro1y e"g" 2APro1y or .y&KG
full $ethod in this custo$er $ust i$ple$ent the functions for storing the
state of the ser%er so that on the ne1t start of the syste$ this state can be resu$ed" 0or repairing the -.
fault sensors can be used" 0irst the faulty -. is deallocated fro$ the 8ob" &econd a new -. is allocated"
Third start the tasks that are running on the failed -." 0ourth restore the state of th
0ault tolerance syste$" The custo$er cannot see all these types of the fault" These can be $onitored
This is done by a $onitoring syste$ co$posed of the sensors" These
A Virtuali5ation and Fault 3olerance Approach
0ault tolerance is pro%ided to cloud infrastructure by i$ple$enting the cloud $anager '1?( Goad
balancer 0ault 2andler and the )ecision $aker" A para$eter success rate is used specially for fault
In this Lob is gi%en to the %irtual $achine which has the success rate $ore than so$e specific
%alue" In this way the chances of the fault decrease" The fault handler has the responsibility that when a
-. is found to be faulty its perfor$ance table $ust be updated" According to the success rate and the
perfor$ance table the cloud infrastructure is $ade $ore fault tolerant"
International Journal of Advance Foundation and Research in Computer (IJAFRC)
Volume 1, Issue 6, June 21!" I##$ 2%!& ' !&(%
---"i.afrc"or,
are $ainly three types of failures: Application failure -irtual $achine failure and 2ardware failure" And
in figure 7:

0irst fault tolerance $ethod concentrates in stateless applications like pro1y e"g" 2APro1y or .y&KGH
full $ethod in this custo$er $ust i$ple$ent the functions for storing the
he syste$ this state can be resu$ed" 0or repairing the -.
fault sensors can be used" 0irst the faulty -. is deallocated fro$ the 8ob" &econd a new -. is allocated"
Third start the tasks that are running on the failed -." 0ourth restore the state of the -. for physical
0ault tolerance syste$" The custo$er cannot see all these types of the fault" These can be $onitored by
This is done by a $onitoring syste$ co$posed of the sensors" These
0ault tolerance is pro%ided to cloud infrastructure by i$ple$enting the cloud $anager '1?( Goad
balancer 0ault 2andler and the )ecision $aker" A para$eter success rate is used specially for fault
In this Lob is gi%en to the %irtual $achine which has the success rate $ore than so$e specific
%alue" In this way the chances of the fault decrease" The fault handler has the responsibility that when a
e updated" According to the success rate and the
International Journal of Advance Foundation and Research in Computer (IJAFRC)
Volume 1, Issue 6, June 21!" I##$ 2%!& ' !&(%
11! * + 21!, IJAFRC All Ri,hts Reserved ---"i.afrc"or,
III" 3A2=?AR A$A?@#I#

Tabular analysis of different approaches is done as shown in table 1" 0irst colu$n tells about the
co$ponent for reliability $easure$ents these co$ponents are like -. so$e approaches broadly
consider the reliability of the whole infrastructure and the syste$" The second colu$n represents the
$ethodology used for cloud reliability $easure$ent" Third colu$n represents the techni+ues and
co$ponents used for reliability $easure$ent" It also represents the effect of techni+ues on the reliability
of the syste$"

#l
$o"
3echni4ue
$ame
Components for
the relia1ilit/
measurement
:ethodolo,/
for
measurin,
relia1ilit/
Relia1ilit/
measurement
=sed for
!"1 Adapti%e fault
tolerance
,A0T/
Reliability of each
%irtual $achine is
$easured"
-irtual
$achines are
di%ided into
two category
running and
ad8udication
%irtual
$achines"
Acceptance
test Ti$e
checker
Reliability
assessor
)ecision
$echanis$
Reco%ery cache
are used"
Real ti$e
cloud
co$puting
!"2 Cloud &er%ice
Reliability
.odelling
and Analysis
Reliability of the
syste$ is $easured
Reliability is
di%ided into
two parts:
re+uest ti$e
reliability and
e1ecution
ti$e
reliability
Total reliability
is $easured by
the product of
the two
reliabilities"
2andling
different
failures in
Cloud
Co$puting
3n%iron$ent
!"5 0aultH
Tolerant and
Reliable
Co$putation
in Cloud
Co$puting
Reliability of the
ser%er is $easured
Reliability
and the cost
relation are
studied"
Reliable
co$ponent
with the
reliability
greater than a
threshold %alue
and ha%ing less
cost is selected"
<eneral
&cientific
co$putation
!"7 0ault
Tolerance
and
Resilience
0ailure of a
$achine
0aults are
di%ided into
two parts
Crash faults
and
!y=antine
0aults"
Reliability of
the syste$
increase with
the checkpoint
restart
replace$ent"
Characteri=ing
recurrent
failures in
Cloud
en%iron$ent
!"; 0ault
Tolerance
.iddleware
0ault in the syste$ 0aults are
di%ided into
two parts
Crash and
Ti$ing fault"
!y pro%iding a
$iddleware
ser%ice o%erall
reliability of
the syste$
increases"
)istributed
applications
fault tolerance




!"> A &yste$
Ge%el
Approach
Reliability as a
ser%ice
!y
introducing
the 0T. for
Reliability of
the syste$
increases by
pro%iding
fault tolerance
property as on
International Journal of Advance Foundation and Research in Computer (IJAFRC)
Volume 1, Issue 6, June 21!" I##$ 2%!& ' !&(%
11( * + 21!, IJAFRC All Ri,hts Reserved ---"i.afrc"or,
reliability using 0T." de$and
ser%ice
!"? 0ault
Tolerant
Approaches
in Cloud
Infrastructure
0aults in the syste$ 0aults are
di%ided into
application
%irtual
$achine and
the physical
node faults"
Reliability is
increased by
the stateless
and the stateH
full approaches"
Replication and
the sensors
increase the
fault tolerance"
Autono$ic
repairing of
faults
!"@ A
-irtuali=ation
and 0ault
Tolerance
Approach

Cloud
infrastructure
0aults are
handled by
the fault
handler"
Auto$atic
updating the
syste$ by the
fault handler
and using
success rate
para$eter to
increase the
reliability of
the syste$"
Reducing the
ser%ice ti$e
and increasing
the syste$
a%ailability in
a Cloud
en%iron$ent


IV" C0$C?=#I0$ A$< F=3=R> E0RD

In this paper we ha%e studied different approaches for cloud co$puting reliability" There are $any
issues about the cloud reliability like the heterogeneity dyna$ic nature etc" Reliability of cloud
co$puting depends on the reliability of its co$ponents like -. Physical nodes or the application
running on the cloud en%iron$ent" There are se%eral types of faults in the cloud en%iron$ent like crash
fault ti$ing fault application faults etc" Reliability of the cloud en%iron$ent can be increased by
replication restart continuous auditing of all the infor$ation about each co$ponent of cloud
en%iron$ent by using efficient sensors for $onitoring" In future we will work on the i$pro%e$ent of the
cloud reliability by proposing a $echanis$ which i$ple$ents a collection of these reliability approaches"

V" R>F>R>$C>#

'1( National Institute of standards and technology *"& )epart$ent of Co$$erce special publication
@EEH17; Peter .ell Ti$othy <rance"
'2( Introduction to Cloud Co$puting architecture white paper sun $icrosyste$"
'5( Gee !adger Ti$ <rance Robert PattHCorner Leff -oas MCloud Co$puting &ynopsis and
Reco$$endationsN NI&T &pecial Publication @EEH17>"
'7( &heheryar .alik 0abrice 2uet MAdapti%e 0ault Tolerance in Real Ti$e Cloud Co$putingN 4orld
Congress on &er%ices 2E11 I333"
';( OuanH&hun )ai !o Oang Lack )ongarra <ewei Phang MCloud &er%ice Reliability: .odeling and
AnalysisN"
'>( )" Abra$son R" !uyya L" <iddy MA co$putational econo$y for grid co$puting and its
i$ple$entation in the Ni$rodH< resource broker" 0uture <eneration Co$puter &yste$sN"
International Journal of Advance Foundation and Research in Computer (IJAFRC)
Volume 1, Issue 6, June 21!" I##$ 2%!& ' !&(%
116 * + 21!, IJAFRC All Ri,hts Reserved ---"i.afrc"or,
'?( O"&" )ai ." Qie R"G" Poh MReliability of grid ser%ice syste$s Co$puters S Industrial 3ngineeringN
;E,1H2/ 15EH17?"
'@( O"&" )ai ." Qie R"G" Poh MReliability Analysis of <rid Co$puting &yste$sNtheAth I333 Pacific Ri$
&y$posiu$ on )ependable Co$puting I333 Co$puter Press"
'A( Ling )eng &cott C"H2" 2uang Ounghsiang &" 2an Lulia 2" )eng M0aultHTolerant and Reliable
Co$putation in Cloud Co$putingN"
'1E( Ra%i Lhawar -incen=o Piuri M0ault Tolerance and Resilience in Cloud Co$puting 3n%iron$entsN"
'11( 4enbing PhaoP" ." .elliarH&$ith and G" 3" .oser M0ault Tolerance .iddleware for Cloud
Co$putingN A?@HEH?>A;H715EH5C1E I333"
'12( Ra%i Lhawar -incen=o Piuri .arco &anta$brogioy MA Co$prehensi%e Conceptual &yste$HGe%el
Approach to 0ault Tolerance in Cloud Co$putingN A?@H1H7>?5HE?;EH; 2E12 I333"
'15( T" 3rl M&er%iceH9riented Architecture: Concepts Technology and )esignN *&A: Prentice 2all PTR"
'17( Alain Tchana Gaurent !roto )aniel 2agi$ont M0ault Tolerance Approaches in Cloud Co$puting
InfrastructuresN I&!N: A?@H1H>12E@H1@?H@ 2E12 I333"
'1;( .icrosoft M4indows a=ure: .icrosofts cloud ser%ices platfor$N
http:CCwww"$icrosoft"co$Cwindowsa=ureC"
'1>( 4alters Lohn Paul Chaudhary -ipin MA faultHtolerant strategy for %irtuali=ed hpc clustersN The
Lournal of &uperco$puting"
'1?( Pranesh )as )r" Pabitra .ohan Rhilar M-0T: A -irtuali=ation and 0ault Tolerance Approach for
Cloud Co$putingN A?@H1H7>?5H;?;@H>C15C 2E15 I333"

You might also like