You are on page 1of 10

REPORT ON SOFTWRAE DEFINED

NETWORKING FOR CLOUD


COMPUTING
Suchitra Jambawalikar

1.

Introduction
In Cloud Computing we use a network of remote servers. These

remote servers are hosted on the internet. This lets us manage, store and
process data in a centralized manner rather than storing to and accessing it
from our personal computers. [1] Cloud Computing, also referred to as the
cloud, allows users to have access to shared resources very efficiently. [1]
Cloud Computing creates a community which is joined by many users. However as the major attraction of cloud computing is sharing resources among
multiple users,this multi-tenancy could create some obstacles. It is very important to make sure that one users data is not affected by another user. Hence
there is a need for partitioning every users resources and keep it protected.
Ethernet and IP can be used to provide isolation but it cannot serve a large
group of users. [2]
This gives rise to the need of Software Defined Networking

(SDN)[5][9]. In SDN the control of shaping the traffic is shifted from the
hardware to the software. Also it becomes easier for the network administrators to handle the traffic load centrally and in a efficient manner. [3]

2.

Challenges in the existing cloud computing


architecture
For now the cloud computing architecture follows the pattern

of everything can be fit into the same architecture. But Cloud has varied
requirements. In the existing architecture the network design, the rules of
forwarding elements and the security detail are implemented by considering
all the requirements. So basically a general architecture that works for all the
cases is implemented. This prevents from acquiring an optimal network usage.
[4]
To achieve optimal network usage it is necessary that the traffic
is isolated and the end users are provided with control access. To implement
these policies the configuration of every switch will have to be changed. Also
each switch will have different requirements and they need to be changed
accordingly. Building such a system for cloud computing will be very complicated. So the best solution in this case is to use Software Defined Networks.
[4]

3.

Role of SDN in Cloud Computing


There are two SDN Models that can be used for cloud comput-

ing, the overlay model and the network model. The virtual network is
created by the software in the overlay model and in the network model it is
created by the network devices.[2]
VMwares Nicira technology is an example of overlay model. An
abstraction layer is created by the Niciras Network Virtualization platform
between the physical network and virtual switches deployed on hypervisor
hosts. In Overlay model the IP addresses are partitioned into multiple virtual
sub networks by using software. These sub networks are then accessed by
network APIs as if they were IP networks. The traffic that is passed over
these sub networks is kept separated and secure by the software. [2]
Network devices are used to build Network-hosted SDNs. In
this model the network traffic is managed directly by the SDN. Some network
vendors expand network devices into simpler form by removing the complicated
traffic management policies.[2]

4.

Road to SDN
SDN in a way lets you program networks. Before SDN came

into picture there were other efforts made to program networks. Programming
networks started with active networking which allowed us to dynamically modify network operations while sending packets. Next came the control and data
plane separation which allowed us to separate the control planes and data
3

planes and also build interfaces between them. Later came the Open Flow
and Network OS, which was the first time that open interface was adopted.
We also found methods to practically separate the control planes from the
data planes.[5] Then finally came the Network Virtualization. [5]

4.1.

Active Networking
Traditionally networks could not be programmed. Active Net-

working let you program networks. You could basically program the data you
send over the network to change its characteristics according to the channel
it is going through. But, this did not agree with people who believed that for
internet to be successful we must keep the network simple.[5]

4.2.

Data Plane and Control Plane Seperation


The messages that are created by the users are included in data

planes. [6] These messages are delivered over the network using L2 forwarding
protocols or L3 routing protocols like spanning tree and Open Shortest Path
first respectively. [6] The messages that control the delivering of these user
messages are known as control planes. The control messages creates the forwarding table whereas the packets needed for forwarding are contained in the
data planes. These packets use the forwarding table to forward packets. This
logic of forwarding the packets by the data plane is executed by the switches
and this logic is very simple to implement, thus reducing its costs. [6]

4.2.1.

Centralized Control Plane


Having a centralized control plane means that it has all the

knowledge about the networks including its topology and where the hosts is
connected to the network. Having a centralized control plane means that the
control plane knows everything that is going on within the network. This helps
the control plane is building simpler forwarding policies as it has an idea of
how the entire network is running. [6]

4.2.2.

Programmable Control Plane


By having a centralized control plane it becomes very feasible

for the network manager to control the network implementing the policies and
changing them dynamically as and when needed. This programmable feature
of the control plane is very important to cloud computing as it is this feature
that lets us divide the networks into multiple virtual networks with each having
different policies. [6] This feature is what lets the multiple cloud users share
a network and still have the privacy that their data will not be modified by
other clod users sharing the same network. [6]

5.

Network Virtulization in Cloud Computing


The interest in Network Virtualization has increased because

of the tremendous rise in Cloud Computing. Network Virtualization can be


considered to be a use case for SDN. Network Virtualization lets you run
many virtual networks at the same time. These virtual networks have a simpler

topology compared to the physical networks. Network Virtualization and SDN


are very closely related. [6] The following are the three ways through which
SDN and Network Virtualization are connected[5]

5.1.

Using SDN as a technology for Network Virtulization


Network Virtualization came to a rise because of cloud comput-

ing so that multiple cloud users can use the same network. One way of letting
multiple users share the same network is to use overlay networking which is
basically a neywork built on top of another network. But each node in the
overlay network though connected to each other through virtual links still is
an extension of a physical network. A better idea is to have a centralized
control plane like in the SDN that controls how the packets are delivered from
one location to another while at the same time making sure the user data is
protected even if they share the same network infrastructure. [5]

5.2.

Using Network Virtulization to evaluate and test


SDNs
Before setting up the SDN over a network, it is possible to

evaluate and test the SDN application on a virtual environment because of its
ability to separate the control plane from the data plane. [5] Having a virtual
environment helps the network manager to test and evaluate the control plane
for a network that has lots of hosts and switches on it on a single machine. [5]

5.3.

Slicing the SDN


It is easier to slice a SDN compared to a router. [5] What it

means by slicing the SDN is that the entire traffic is divided into slices and
each of these slices will be provided with a percentage of network resources
and each of these slices will be controlled by different SDN. [5]

6.

Standarized API
The centralized control plane along with southbound API and

northbound API together forms the SDN. [6] Southbound API is used to communicate with the hardware and the northbound API is used to communicate
with the network. The control plane is then split into a hypervisor layer and
control system layer. [6] This hypervisor layer is used as an intermediary
between the forwarding hardware and the controllers. [6]
Southbound APIs allows the SDN controller to make changes
in the network according to real time needs. [7] Open Flow is an example
of Southbound API. It is also the most famous Southbound API and is standardized by Open Network Foundation. [6] It lets researches run experimental
policies on the network. It also provides a way in which the controller should
interact with the data plane. [8] The entries in the forwarding table can be
added and removed depending on real time traffic, in OpenFlow. [8] [6]
Northbound APIs are not yet standardized. Here every controller can have separate programming interface. [6] It is very important that
the APIs are standardized otherwise the SDN development will be very lim-

ited. Also having east west API will be very helpful as it will allow controllers
from adjacent domains to communicate with each other. [6]

7.

Flow Based Control


In todays world the file sizes are growing exponentially. But

this has not mean that the packet sizes have also increased. They still remain
the same. So in order to transfer the whole file, you have to send a sequence of
packets. It is observed that you just have to make the control decisions for the
first packet of the sequence. For all the other subsequent packets, the same
control decision will be reused. This therefore reduces the interaction between
the controller and the forwarding item. [6] The forwarding item requests the
control information when it receives the first packet of the sequence. The
mask of the packet headers helps in determining the sequence of the packets.
[6] Also the forwarding table has the details regarding how to manage packets
with matching headers. [6]

8.

Conclusion
Cloud Computing was born because of all the improvements

made in virtualization. We now store data virtually, we also have virtual


networks. A lot of progress has been made in virtualizing Networks. One
of the biggest contributor to the advancement in Network Virtualization is
Software Defined Networks. Cloud users have also accepted SDN. [2] SDN also
helps in reducing the cost and greatly helps in solving the problems caused by
8

multi-tenancy. [6]
It is believed that in future SDN will make networks easily
virtualizable. [6] Having virtualized networks is very important for cloud computing where multiple users share the same network. Also considering how
easy it is to implement the data planes the forwarding elements will be very
affordable. Thus SDN is more than likely going to reduce the capital required
for cloud computing. [6]

REFERENCES
[1] Vangie Beal. cloud computing (the cloud).
[2] Margaret Rouse. Software defined networking.
[3] Tom Nolle. The role of software-defined networks in cloud computing.
[4] Siamak Azodolmolky, Philipp Wieder, and Ramin Yahyapour. Sdn-based
cloud computing networking. In Transparent Optical Networks (ICTON),
2013 15th International Conference on, pages 14. IEEE, 2013.
[5] Nick Feamster, Jennifer Rexford, and Ellen Zegura. The road to sdn: an intellectual history of programmable networks. ACM SIGCOMM Computer
Communication Review, 44(2):8798, 2014.
[6] Raj Jain and Subharthi Paul. Network virtualization and software defined
networking for cloud computing: a survey. Communications Magazine,
IEEE, 51(11):2431, 2013.

[7] SDN Central. What are sdn southbound apis.


[8] Nick McKeown, Tom Anderson, Hari Balakrishnan, Guru Parulkar, Larry
Peterson, Jennifer Rexford, Scott Shenker, and Jonathan Turner. Openflow: enabling innovation in campus networks. ACM SIGCOMM Computer
Communication Review, 38(2):6974, 2008.
[9] Marc Mendonca, Bruno Astuto A Nunes, Xuan-Nam Nguyen, Katia
Obraczka, and Thierry Turletti. A survey of software-defined networking: past, present, and future of programmable networks. hal-00825087,
2013.

10

You might also like