This progress report summarizes work on a final year project focusing on load balancing in data center networks using software defined networking. The report outlines the need for new networking paradigms to handle changing traffic patterns. It discusses SDN and how it separates the control plane from the forwarding plane. The work done so far includes experimenting with SDN controllers and traffic patterns in a virtualized fat tree network topology. Next steps are to develop an automated traffic generation tool and evaluate different controller options.
Software Defined Networking (SDN): A Revolution in Computer Network
Load Balancing in SDNs with Fat Tree Topology
1. PROGRESS REPORT FOR FINAL YEAR PROJECT
BROAD AREA: SOFTWARE DEFINED NETWORKING
SPECIFIC FOCUS: LOAD BALANCING IN DATA CENTRE NETWORK
By: Aditya Shukla, Anshuman Singh
Guide: Mr. Nabajyoti Medhi
2. Outline
Focus
Need for a new Network Paradigm
Limitations of Current Networking Technology
Basics of SDN
Work done
Next steps
References
3. Need for a new Network
Paradigm
Changing traffic patterns
The “consumerization of IT”
The rise of cloud services
Big data means more bandwidth
4. Limitations of current network:
Complexity that leads to stasis
Inconsistent Policies
Inability to scale
Vendor dependence
5. What is SDN ?
Definition : The physical separation of the
network control plane from the forwarding
plane, and where a control plane controls
several devices.
SDN is a step in the evolution towards active
and programmable networking.
SDN allows network administrators to have
programmable central control of network traffic
via a controller without requiring physical
access to the network's switches.
6. Communication between both
planes:
Openflow: OpenFlow is a programmable
network protocol and associated hardware
designed to effectively manage and direct
traffic by decoupling control and forwarding
layers of routing.
The OpenFlow protocol enables globally-aware
software controllers, which may be
centralized or distributed, to drive the network
edge hardware in order to create an easily
programmable identity-based overlay on top of
the traditional IP core.
7. Focus
To design a mechanism using which we could
load balance traffic in data centre network.
8. Work done
Hands-on practice in Network Virtualization
tool: Mininet & some tools which comes with it
like:
Wireshark: Packet analyser
Iperf: Data throughput analyser
Learned how to operate Floodlight controller
remotely & to connect it with a virtual network.
9. Work done
Experimented with different combinations of
legacy & openflow switches in fat tree network
to figure out best possible combination for
efficiently handling data traffic.
Fat tree Network : is a popular data center
network topology.
It is typically a 3-layer hierarchical tree that
consists of switches on the core, aggregation
and top-of-rack(ToR) layers. The hosts
connect to the switches on the ToR layer.
12. Next Steps
We are now trying to develop a module for
iperf open source tool so as to generate
traffic between all possible pairs of hosts
automatically.
To try different types of controllers on the same
fat tree network so as to figure out most
suitable one for efficiently handling the traffic.
13. References
●Software Defined Networking- The new norm in Networks
https://www.opennetworking.org/images/stories/downloads/sdn-resources/white-papers/wp-sdn-newnorm.
pdf
●About Openflow
http://archive.openflow.org/wk/index.php/OpenFlow_Tutorial
●Mininet tutorial
http://mininet.org/walkthrough/
Editor's Notes
Decoupling
Extract out control from devices.
Network intelligence kept at a place . One logic entity.
Abstraction of underlying network provided to applications.
Simplifies.
Programmability
Real time behavior change
Open API between planes. SDN makes network “applicationcustomized rather than application aware. And makes application network capability aware.
Both side of interface.
Allows network to be programmed per flow basis so more granular control.