Playback-Buffer Equalization for Streaming Media using Stateless Transport Prioritization

Streaming media applications are plagued with playback disruptions due to client buffer underflow resulting from fluctuations in network throughput. When some clients are experiencing buffer underflow and disruptions, other clients may have excess buffered data. This suggests that disruptions can be reduced by the redistribution of resources to provide higher throughput to streaming sessions with less buffered data, that is, to sessions with greater need. This paper proposes a scheme that achieves an effective or end-to-end result of the above dynamic resource allocation without the use of either explicit per-session resource reservation or across-session coordination. Instead, each packet is stamped at the source with the buffer occupancy of the session, and prioritized scheduling is performed at the resource bottleneck to favor packets with smaller buffer occupancy. Due to the actions of end-to-end congestion control, higher throughput can then be maintained for streaming sessions with less buffer. Experimental and simulation results show that the proposed scheme can increase the mean time between playback disruptions by a factor of 4 for fixed initial buffer time, or reduce the required initial buffer by a factor of 2.5 for fixed probability of hitting a playback disruption.