Research of Packet Classification on Network Protocols by Entropy Theory

Packet classification is a key function for software defined networking switches. We study the method of packet classification in network protocol, and our goal is to match forwarding rules according to several fields in the head of packets and the rule database. On the basis of the existing string matching model, a Key-Position Vector (KPV) algorithm is put forward, which aims to find some key positions for partitioning all the rules into evenly subsets. Some mathematical background of this problem is provided first, and then the joint and conditional entropy theory is applied to determine the KPV positions step by step in order to achieve an approximately optimized KPV for reducing the searching subsets' size. Moreover, this model possesses important significance in improving search efficiency, and an example is given to illustrate it at the end of this paper.