This award is funded under the American Recovery and Reinvestment Act of 2009 (Public Law 111-5).
Although there are several large-scale industrial deployments of peer-to-peer (P2P) live video systems, these existing systems have several fundamental performance problems, including huge channel switching delays, large playback lags, poor performance for less-popular channels, ISP unfriendliness. In these traditional systems, a peer only redistributes the video it is currently watching. In this research, the PIs are exploring a radically different approach to P2P live video streaming, View-Upload Decoupling (VUD). The main idea of VUD is to have each peer distribute one or more channels, with the assignments being made independently of what the peer is viewing. This novel approach has three major advantages over the traditional isolated-channel designs: channel-churn immunity; cross-channel multiplexing; and the enabling of structured streaming. The PIs are developing tractable analytical performance models for multi-channel P2P video streaming systems, for both VUD and traditional design approaches. The analytical results not only highlight the advantages of the VUD approach, but also provide important ``rules-of-thumb'' for the design of VUD systems. The PIs are developing dynamic VUD provisioning algorithms that are both robust with respect to channel churn and also adapt to dynamic channel popularity and flash crowds. The PIs are developing VUD provisioning, management and streaming schemes that take into account ISP locality and largely reduce the video streaming traffic imposed on ISP networks. The PIs and their PhD students are also developing an open-source VUD prototype.
