摘要
提出了一种适用于比例区分服务的数据包标记算法———基于速率预测的三色标记算法(RatePredictionMarker,RPM)。RPM算法基于在线流量预测算法,对DiffServ网络中经过分类器划分的网络流或者进入标记器的网络流进行提前1期流量预测,根据预测结果以及历史流量的均值进行加权后的结果对数据包进行标记。当源端可能进入TCP的拥塞恢复阶段时,按比例将网络中的可用带宽分配给服务聚集流,从而提高网络带宽的利用率。仿真和实验结果表明,RPM算法与现有的算法相比在目标速率影响和带宽吞吐量上具有更好的性能,从而获得了很好的公平性和带宽利用率。
A new algorithm of data packet marker named three color marker algorithm based on rate prediction （ Rate Prediction Marker, RPM） which fit for differentiated service was proposed. RPM algorithm is based on online traffic predictors algorithm. Network flows in DiffServ network were either divided by classifier or entering into marker makes flow prediction ahead of one period by RPM algorithm based on online traffic predictors algorithm. The packets were marked according to the results of prediction and the average weighed value of history flows data. When the sources may enter the recovery period of TCP congestion, the network available bandwidth was proportionally designated to serve congregation flows, and this raised the network bandwidth utilization rate. Simulation and experiment＇s results show that RPM algorithm has much higher performance when influencing aim rate and bandwidth＇s throughout compared with the existing algorithms, and get fairly fairness and bandwidth utilization rate.
出处
《计算机应用》
CSCD
北大核心
2005年第12期2888-2891,共4页
journal of Computer Applications
关键词
流量预测
公平性
标记算法
区分服务
traffic prediction
fairness
marker algorithm
differentiated service
