Download files
Abstract
Video traffic is increasingly dominating broadband networks, creating new challenges for Internet Service Provider (ISPs) and Video Content Providers (VCPs) alike. ISPs are struggling to monitor and manage bandwidth usage of video traffic so it can coexist with other traffic types, while VCPs are striving to maintain good user experience in the face of changing network conditions. This thesis is an exploration on the use of Software Defined Networking (SDN) technology to provide fine-grained visibility of video traffic on ISP networks, while enhancing network state visibility for VCPs to better adapt their video transmission schemes.
We begin this thesis by surveying various methods for network traffic classification and telemetry, and existing techniques for adaptive video transmission based on estimating network state. We then present the first major contribution of this thesis - the development of a tool called TeleScope that ISPs can use to identify streaming video flows from various sources such as Netflix and YouTube, and characterise them in terms of end-points, duration, and bit-rate at very low cost using off-the-shelf SDN switches. Our tool is prototyped using commodity hardware and open-source software, evaluated for performance and scalability in a lab setting, and deployed in a campus dormitory network to reveal new insights on video viewing patterns. Our second major contribution proposes and implements a framework for the network to reveal real-time state information to interactive video applications using explicit APIs. Using these, we design and implement congestion control and rate slicing algorithms that interactive applications and the network can cooperate on to deliver at low response latency while achieving high link utilisation. This thesis demonstrates that the use of SDN technology can enhance video traffic delivery and management to the benefit of both ISPs and VCPs.