校园网流量多维度行为分析及优化

Multi-dimensional behavior analysis and optimization of traffic in campus network

针对校园网出口链路拥塞压力不断增长,粗粒度的流控策略难以适应变化中的校园网流量环境的问题,分析了目前校园网出口高带宽占用比的几种网络协议的流量特征,从时间、区域、接入、链路多个维度对校园网的总体流量行为进行了深入的分析,根据分析结果提出了更细粒度的流量优化策略.结果显示:校园网入流量中HTTP流媒体(HTTP Streaming)流量占比最高且增长迅速,用户入带宽分布呈幂律分布特征,不同的区域、接入方式、ISP链路对网络应用协议的偏好度不同,更细粒度的流量优化策略可以有效缓解校园网出口拥塞,优化网络性能.

Aimed at increasing congestion in campus network export and the coarse-grain flow control strategy adapting to the changing campus traffic environment difficultly,traffic features of several high bandwidth occupied network protocols were analyzed,general traffic behavior in campus network was analyzed deeply from time,regional,access,and chain multiple dimensions,more fine-grained traffic optimization strategies were proposed according to the analysis results.The results show that,HTTP Streaming traffic takes the highest percentage of campus inbound traffic and grows quickly,bandwidth of users inbound traffic appears power law distribution,different regions,access ways,and ISP chains have different preferences on network application protocols.More fine-grained traffic optimization strategies can ease campus network export congestion and optimize network performance effectively.