Network Embedded FEC for Overlay and P2P Multicast Over Channels with Memory

4 5 6 Forward Error Correction (FEC) has been used extensively for distributing multimedia contents over the internet. Traditionally, FEC can only be implemented in an end-to-end mode. In [3] we have introduced network embedded FEC (NEF) framework and showed that NEF can greatly improve the decodable probability and message goodput in overlay and peer-to-peer (p2p) multicast networks. In NEF, FEC encoders and decoders (codec) are placed in the intermediate nodes of a network; the NEF codecs detect and recover lost packets within FEC blocks at earlier stages before these blocks arrive at deeper intermediate nodes or at the final leaf nodes. Analysis in [3] is based on the assumption that packet losses follow a binomial distribution. In practice, packet losses are often correlated and occur in bursts. In this paper, we evaluate the performance of NEF within random multicast distribution trees that exhibit losses-with-memory over their branches. First, we derive a closed expression for the probability that a node receives i packets as the sender sends n packets when the packet losses follow a Gilbert model. Based on this analysis, we evaluate the impact of various loss burst lengths on the performance of NEF within random multicast trees. Results show that even with very high packet loss correlation, NEF outperforms end-to-end FEC and can improve decodable probability and message goodput dramatically.