Enhancing the performance of multi-source streaming system using the MCMCF-based algorithm

We present the implementation and evaluation of an MPEG-4 multi-source streaming system, in which each receiver can simultaneously retrieve different parts of a streaming object. Unlike the traditional single-source streaming system that may have poor perceived quality at the receiver site due to the network congestion or the heavy-loaded server, our multi-source system aims to maximize the usage of available resources on both end-hosts and the network to provide better quality for the users. In order to achieve this objective, our system employs three key components: a) select appropriate servers, b) distribute load to those selected servers to maximize the overall throughput, and c) avoid self-congestion between different object flows that are requested by different users at the same time. We apply a minimum-cost multicommodity flow (MCMCF) (J. Castro and N. Nabona, 1996) based approach to deal with the above issues. We have extensively evaluated the performance of our system and other different streaming systems through the simulation. The simulation results show that our system can achieve better end-to-end throughput than greedy-based or min-cost-flow-based streaming systems does.