MSc Graduation project software defined networking
The myriad of applications running in a data center (DC) demands for an highly efficient and agile management and control plane to optimize on-demand the DC resources (servers, switches, and other network resources). Software defined networking (SDN) and network function virtualization (NFV) are promising techniques to dynamically virtualize, control, and optimize the DC resources allocation based on the running applications. To implement the SDN and NFV control, all the hardware elements of the DC (servers, electrical and optical switches) should be equipped with an interface (OpenFlow agent) to allow the SDN to fully control and monitor the overall DC hardware. Thus, the SDN can dynamical provision multiple virtual networks, migrate virtual machines on demand, load balancing the traffic, and more functions by properly (re-)configuring the network elements and the servers. Figure 1a shows a typical data center architecture with servers aggregated in racks, top-of-the-rack (ToR) switches, the DC network switches, and on top the SDN control plane. Figure 1b shows the OpenFlow Agent interface and the SDN control plane management GUI (OpenDayLight) to be developed in the project.
The project main objective is to develop the OpenFlow agents to interface the SDN controller and the optical switch and ToRs controllers based on FPGA. The assignment comprises the following main tasks:
1. Develop in Java/C or VHDL the optical switch OpenFlow agent embedded in the FPGA based optical switch controller. Implement the optical communication with the SDN controller (OpenDayLight management GUI) via a 1 G Ethernet optical transmission.
2. Develop the ToR OpenFlow agent embedded in the FPGA based ToR switch, and implement the communication with the SDN controller similarly as in point 1.
3. Develop the SDN controller API for the dynamic (re-)configuration of multiple virtual networks by provisioning the look up tables and the configuration of the optical switch and the ToR .
4. Assess the DC performance under DC real applications and monitoring the network statistics (average load, packets loss, latency) by state-of-the-art network performance analyser.
5. Develop API at the SDN controller employing the monitored statistics for load balancing and intelligent scheduling algorithm for resource allocation and performance optimization.
More Information: Dr. Nicola Calabretta, Flux 9.087, Email: email@example.com
PhD Candidate Xuwei Xue, Email: firstname.lastname@example.org