Impacts of Using Programmable Network Architectures on Datastreams

Abstract: With an increase of popularity in live video streaming, there are new demands on current networks. Therefore it is interesting to explore if new architectures has what it takes. Software defined networking (SDN) is a new architecture that brings benefits in terms of control and management of networks. In this thesis we explore SDN's capabilities in terms of live streaming and put forth the question: What are the effects of using SDN-based network architecture on data streams with respect to jitter and latency? To answer this question we also need to know how to design an evaluation framework for measuring jitter and latency. We developed two iterations of a live stream, time stamping each video frame. These were used in two similar network topologies, a conventional network and an SDN-network using OpenFlow. The first iteration of the live stream depended on clock synchronization. After inconsistencies between tests we suspected that the clock synchronization was not accurate and reprogrammed the stream. The second iterations showed the same inconsistencies in latency meaning our results were inconclusive due to randomness in operative systems and inconsistencies in time stamping. This work shows several considerations to take into account when measuring latency on an application level in a network environment. Our most distinctive result shows that SDN and conventional networks have comparable jitter.