Video Splicing Techniques for P2P Video Streaming

In HTTP live streaming (HLS), a video is spliced into multiple segments of equal duration. The size of the segments has a major impact on the stalls. In adaptive video streaming, clients adapt the bit-rate of the video to minimize the stalls. But it often degrades the video quality. In order to maintain a high video quality, we can adapt the duration of the segments instead of the bit-rate. Thus the splicing technique can have a major impact on a video streaming. In this paper, we studied different video splicing techniques for peer-to-peer video streaming. We spliced MPEG-4 video based on different durations and Group of Picture (GOP). We found that the duration based splicing performs better than the GOP based splicing, though the duration based splicing requires much more data to be transferred than the GOP based splicing. In P2P video streaming, peers can leave the swarm anytime. To maximize the availability of a segment, peers often download multiple segments simultaneously. Again, if a peer downloads too many segments at a time, it will increase stalls. In this paper, we proposed a formula to calculate the number of video segments a peer should download simultaneously in order to reduce the stalls. We conducted the experiment on GENI.