On Optimal Scheduling for Layered Video Streaming in Heterogeneous Peer-to-peer Networks
Xin Xiao,Yuanchun Shi,Yuan Gao
DOI: https://doi.org/10.1145/1459359.1459486
2008-01-01
Abstract:Layered video streaming in peer-to-peer networks has drawn great interests since not only it can accommodate a large number of users but also it handles heterogeneities of client networks. However, to our knowledge, there's still a lack of systematical study on the data scheduling (i.e. requesting and relaying data) for layered streaming, and previous works in this area just focus on maximizing the throughput and/or minimizing the packet delay. In this paper, firstly, according to the characteristics caused by layered coding, we propose the four objectives that should be achieved by data scheduling for layered streaming: high throughput, high layer delivery ratio, low useless packets ratio and low subscription jitter; then, we use a 3-stage scheduling approach to request missed blocks, where each stage has different scheduling objective but collaborate with each other. The min-cost network flow model, probability decision mechanism and multi-window remedy mechanism are employed in Free Stage, Decision Stage and Remedy Stage, respectively, to achieve the above four goals. Extensive experimental results indicate that our approach outperforms other schemes in both throughput and layer delivery ratio. Besides, the useless packets number and subscription jitters are kept low.