Stream Classification Algorithm Applied Research In High-speed Routers

For extraordinary rapidly development of the Internet at present, the service providers have to apply themselves to higher performance network. With the constantly improvement of skeleton Internet speed, differential service has been the research hotspot of network. In order to provide the network status guarantee of the bandwidth, delay, the ratio of packet lost etc., Internet should classify all kinds of packet for different requests.This thesis studies profoundly the actual packet classification algorithms at first. It analyzes the solution to packet classification by selecting some representative algorithms respectively, such as Grid of Tree (GoT), Aggregated Bit Vector (ABV), Recursive Flow Classification (RFC), and Tuple Space Search (TSS). The article compares these algorithms systematically in several aspects: basic ideology, thread to carry out, actual environment, capability and so on.In the multi-application integrated environment, there are many requests of packet classification, including Policy Based Routing (PBR), Network Address Translation (NAT), Reversed Path Forwarding (RPF), Quantity of Service (QoS) and so on. On the basis of the analysis the actual network environment of High-Speed Router and the research about main solution of that, this thesis carries out an overall design of the problem of packet classification in High-Speed Router, it applies to share the rule library and to find the best combinative point. In the end, this thesis designs an emulation system, it also makes statistic about many data of packet classification problem, such as the space occupation, the preprocess time and the packets filter time. The result makes out the effiective performance of the design.