面向高速对等网应用的拥塞控制机制

The Congestion Control Scheme Aiming at P2P Applications in High-Speed Networks

为改善现有拥塞控制机制在高速网络中带宽利用率偏低、稳定性不高和更好地应对急剧增长的对等网(P2P)应用流量,提出了一种以FAST为基础,应用大规模网络测量和模糊控制技术的拥塞控制机制.该机制利用分布在网络中的测量设施周期性获取网络状态信息,指导端系统选择适当的控制参数.单个和多个链路瓶颈条件下的仿真实验均表明,本文拥塞控制机制能够在高带宽时延积网络中获得更高的带宽利用率和稳定的排队时延,很适合于P2P等数据传输量大、连接持续时间长的流量的拥塞控制.

To account for the low utilization and oscillation problems of the existing congestion control schemes under high-speed networks, and to meet the quickly increasing P2P traffic, a novel congestion control scheme based on FAST was proposed, which employs large-scale network measurement and fuzzy logic control techniques. The main idea of this scheme is to gather network performance data periodically, by exploiting measurement facilities distributed in networks, to help end systems determining appropriate control parameters. The simulation results in high bandwidth delay product networks indicate that this scheme can achieve higher throughput and more steady queuing delay, independent of the number of bottleneck links. It is particularly suitable for P2P like traffic with high volume, long-lived flows.