Network Load-aware User Grouping for Internet Media Streaming Systems

While emerging Internet media streaming systems promise to offer viewers an abundant set of user interactivities and controls over the contents, providing streaming services through a large number of concurrent point-to-point connections stresses both server and network because of the large volumes of data and relatively high bandwidth requirement. While the server limitation can be circumvented by deploying server clusters, the network limitation is far less easy to deal with, due to the difficulty in measuring and balancing network load. Therefore, successful deployment of media streaming services requires a scheme that is conducible to mitigate this problem. In this paper, we employ two network load metrics, the worst link stress and the degree of interference, for the measurement of the network load balance. Then, we formulate the user grouping problem and present a greedy algorithm that reduces the network load metrics for each session and also well balances across the sessions. Through the simulation results, we conclude that our algorithm perform better than the existing scheme in the following aspects: (1) load-balancing in the network, (2) total bandwidth used by the connections.