Research On Packet Classification Method Based On Network Processor

With the development of a variety of network applications,packet filtering,differentiated services,quality of service,multicast,and meter fees require packet classification.So classification determines the performance of the network applications.Network processor has the flexibility of general purpose processor and high performance of ASIC.It is considered to push forward the development of next generation network core equipment.Firstly,the paper studies the packet classification method definition,criteria and design principles.And linear packet classification,parallel packet classification,Trie-based packet classification and Tuple space packet classification has been discussed and analyzed.Secondly,the paper studies with high performance,flexibility,scalability and strong high-end IXP network processor-IXP 2850 system platform.According to the network processor hardware and IXA software framework,some problems which limit the performance of packet classification have been analyzed,such as fast storage resources is lask is poor in IXP 2850.Some typies of suitable packet classification algorithms also have been discussed.Thirdly,according to the research of packet classification and network processors,a new method which is based on parallel bit packet classification is put forward-TBV(Tuple Bit Vector)method.This method use hash unit and micro engines to acquire fast computing and parallel processing,which has better performance than similar methods.Finally,modules in software are developed on the IXP2850 hardware platform,TBV are achieved in IXA framework and some tests are conducted.