Data center ethernet : data traffic congestion control

• Main Author: Ang, Yu Bin
• Other Authors: Chng Eng Siong
• Format: Final Year Project
• Language: English
• Published: 2013
• Subjects: DRNTU::Engineering::Computer science and engineering::Computer systems organization::Computer-communication networks
• Online Access: http://hdl.handle.net/10356/52513
• Institution: Nanyang Technological University

Data Centre Networks have begun to grow in sizes and utilizations. To provide quality services to users, the efficiency and reliability of the interconnected links is of great importance in running a Data centre. In ensuring that quality of service (QOS) is met, congestion control has become one of the critical topics being tackled. In this project we will focus on two different areas in our efforts of congestion control. First we will be working on searching of a network simulator which allows us to run the current OpenFlow based congestion control protocols as well as future development of the protocol. Secondly, we will be constructing an actual Openflow Network to run performance test Equal-Cost multi-path as well as Edge to Edge Lowest cost routing with and without congestion re-routing. The results of both test will be compared to see which algorithm fares better. In the search for a suitable simulator, we discover NS-3 network simulator which is able to intergrade with the Click modular router. The integration of both system, named NS-3-Click, allow us to us Openflow Click element to simulate an OpenFlow network. This configuration gives us the ability to future expands the functionality of an OpenFlow switch if needed. However we met with an incompatibility issue with OpenFlow Click element. We have concluded that it is possible to resolve the incompatibility; unfortunately we do not have sufficient time to resolve the issue. For the performance test, it was discovered that for busty traffics, both routing algorithm performed equally and there is no difference in performance with or without congestion re-routing. On the other hand, ECMP performed significantly better than lowest cost algorithm when no rerouting is implemented as well as when rerouting is implemented for both itself and lowest cost with non-busty traffic.