Dissecting the performance of Live Mesh-based Peer-to-Peer Streaming

Mesh-based Peer-to-Peer streaming (MPS) mechanisms incorporate swarming content delivery and thus are able to support scalable streaming of live content. Their key component is a packet scheduling scheme at each peer that determines pulled packet from neighbors while accommodating in-time arrival and diversity of delivered packets. Besides packet scheduling scheme, the overall performance of a MPS mechanism also depends on the availability of excess resources in the system. Recently proposed MPS mechanisms have been evaluated in a scenariorich setting. Thus, neither their performance in a resource constraint scenario nor the separate effect of packet scheduling and available resources on their performance is known. In this paper, we dissect the performance of MPS mechanism and investigate the effect of packet scheduling and available resource on their performance. We argue that the global pattern of content delivery primarily determines behavior of a MPS system. We present a new evaluation methodology that properly captures this pattern. Using our evaluation methodology, we examine the performance of representative scheduling schemes and the role of available resources. Our findings provide useful insights in design and evaluation of MPS mechanisms.