Congestion Control in Networks with Mixed IP and P2P Traffic

摘要

Services such as multimedia, VoIP, video-conferencing, social networking and others impose new requirements on providers and constraints on network designers. Fair Queueing algorithms like CSFQ or Stochastic Fair BLUE have been used to improve the quality of the packet transmission. Such mechanisms usually supervise the bandwidth consumption per-flow and become helpless in the presence of P2P traffic. In quest for high quality transmission, multimedia applications are designed to use more and more P2P paradigms. As P2P traffic is also exposed to congestion, few works address congestion control in mixed traditional IP (for short called IP traffic) and P2P traffic. In this paper, we propose a model flow for the mixture of the two and present a principle and a method based on per-subscriber flow control, for congestion control. An architecture based on the Token-Based Traffic Control for P2P applications is introduced. The token resource consumed by each subscriber is counted and controls for both core and edge routers are generated in the case of IP and P2P traffic. The traffic is measured at core routers and the measurement data is conveyed to edge routers. They label the Token-Level on incoming packets according to the congestion index, and police the total input token of each P2P subscriber. Simulations results and the analysis of the impact on the performance of this approach on some P2P experiments are given.