Performance Analysis of Buffer management for live streaming in P2P VOD systems for mesh based non-cluster and cluster environment

Abstract This paper explores mesh based clustering for different start video streaming in P2P systems and estimates the performance of non cluster and clustered models. These models are based on mesh based topology of P2P streaming consisting of peer join/leave. A new approach by way of “clustering” peers has been proposed to tackle P2P VOD streaming. The proposed models were simulated and verified using OMNET++ V.4. A mathematical model has been formulated to evaluate the performance metrics such as buffer size, buffer exchange period over frame loss ratio, startup delay and end to end delay. The results show that(i) unpunctual viewers can be enabled view missing part of video from beginning during live streaming, (ii)the associated buffer in conjunction with join/release time provides constant buffer size and thus ability to provide better service to viewers,(iii)Starting and play out delay gets bounded as the number of peers predefined,(iv) clustering minimizes startup delay, end to end delay and frame loss as compared to existing systems, which are important parameters of a P2P system for streaming.