一种基于抖动测量的细粒度拥塞控制机制

A Fine Granularity Congestion Control Mechanism Based on Jitter Measurement

传统的基于RTT的拥塞控制方法存在着链路的不对称性,不能准确地反映单向链路上的网络状况问题。另外,传统的“加性增、乘性降”拥塞控制算法,粒度太粗。针对这些问题,文章提出一种基于抖动测量的细粒度拥塞控制机制,根据测量的抖动值,动态调节预测算法中的琢因子,使之自适应网络状况。同时将网络状态划分五级,对不同级下的网络状态进行不同的控制。实验证明,该算法可及时、有效地控制网络拥塞。

The asymmetry of paths in the traditional Congestion control based on RTT (Round Trip Time) can not get an in-precise result about network state of one way path. Besides this, the control granularity in the traditional 'additive increase, multiplicative decrease' algorithm seems rather coarsely. Considering of these problems, a fine granularity congestion control mechanism based on Jitter measurement is proposed in this paper. In this scheme, according to measured jitter, the ?琢 factor used in the prediction algorithm can be adjusted dynamically to adapt different network states. In the meantime, the network states are partitioned into 5 levels and different control is granted to the network state in different level. Experiment results presents that this scheme is very timely and effective.