基于IXP1200的快速报文分类算法的设计与实现

Design and Implementation of Fast Packet Classification Based on IXP1200

通过对现有报文分类算法的分析和性能比较,并结合分类规则所具有的特性提出了一种新的基于IXP1200网络处理器的多维报文分类算法,称为PCBNP(packet classification based on network processor),并达到了报文的线速转发.算法除了通过减少分类的规则数和分类的域宽来加快分类的速度外,还采用重定向排序索引、位向量表示匹配规则等技术来加快分类的速度,特别是利用了规则的动态分布规律来确定查找报文字段的顺序,通过先查找“分布最均匀的字段”来达到在所有的字段被查找之前提前找到报文匹配的过滤规则的目的.算法具有高速、多维和可扩展的特性,与现有的算法比较,该算法在综合性能上优于已有的报文分类算法.

By analyzing the existing algorithm and the characteristic of a classifier, a fast five dimension algorithm of packet classification based on the IXP1200 network processor is proposed, which gains wire-forward rate. The algorithm makes most use of the dynamic distribution characteristic of a filter rule and it has the characteristic of fast speed and is easily extensible to a large classifier.