面向新型业务的多维快速包分类算法研究*

网络中新兴的新型业务例如P2P，VoIP，Worm等日益增多，在整个网络流量中占有越来越多的比重。对这些新型业务的监测和控制要求路由器等核心设备必须有能力对数据包进行快速和准确的分类。本文从这些新型业务包分类规则库的特点着手，提出在多维多模式匹配情况下的三种包分类决策树，通过实验说明这些决策树应用在新型业务的包分类上与传统的串行决策树性能各自的差异，进而得出适合新型业务包分类的最佳算法。

New service-oriented multi-dimensional fast packet classification algorithm

A variety of new services such as the emerging P2P, VoIP, Worm are growing on networking,and occupy more and more weight of the entire network traffic.On these new services ,requirements for monitoring and control of the core router equipment must be capable of packet classification for fast and accuracy.Study the rule database's characteristics of these new services', proposing three packet classification decision trees with multi-dimensional pattern matching.The experiment shows the performance differences used in the threee new decision trees with the traditional serial decision tree. Then give the best packet classification algorithm for new services.