Optimizing video streaming against transient failures and routing instability

In addition to network congestion, a link/node failure is another major cause of performance degradation for video streaming over the Internet. Such failures may be followed by a long routing instability period, during which packets can be black-holed due to invalid paths or caught in routing loops. This paper proposes a routing proxy approach to improve media streaming adaptation against both link/node failures and network congestion. In particular, we first argue that it is important to distinguish between network performance degradation due to network congestion versus link/node failures, and then model link/node failures using empirical models derived from measurements. We then show how by means of proper congestion control, such timely notifications from the network layer can be exploited at the streaming server to improve the performance of a rate-adaptive automatic retransmission request (ARQ) video streaming scheme. Simulation results show that a rate-adaptive streaming scheme using feedbacks from our proposed proxy can recover much faster from link/node failures than a scheme without such feedbacks.