The topic of my thesis is an OpenFlow testbed creation where I can test a Bloom filter based routing mechanism. The OpenFlow Release separates the the internal operation of switches (forwarding) and the control logic which becomes programmable. The hardware component, which is responsible for the packet forwarding is solved by every single vendor, because it is not standardized. The programmable logic is a major step forward, because the vendors of the devices hid the details of the internal implementation of the switches, thus it was difficult to the researchers to try out the new innovative ideas, new network protocols, routing algorithms in real conditions. At the beginning the OpenFlow-enabled devices were just software implementations, which were running on PCs (OpenFlow Reference Switch and the Open vSwitch), and later hardware devices appeared on the market, that supports the OpenFlow protocol. This kind of devices are can be found in our laboratory: TP-Link switches and a HP ProCurve Switch. Original factory settings do not support the release.
The tested algorithm of my thesis is non-IP routing based, source routing (the route is predetermined in the source) mechanism, which implements the simple Bloom filter to decide where the packets are forwarded.
I planned three different types of networks to test the non-IP routing mechanism. The first topology consists of software switches, the second consists of hardware switches and the third consists of both software and hardware switches.
After the planning, I implemented the networks from the devices, which are available in the laboratory. I analyzed different parameters of the created networks: bandwidth, jitter and latency. I made a detailed documentation about the results of the analysis, and arranged the measurements data into tables, then depicted on diagrams to show the differences between the cases and I made conclusions of the advantages and disadvantages of the algorithm.