Design And Implementation Of Application Layer Packet Classificatory Collector

With the rapid development of Internet, various net traffics are emerging, and this requires the network device to provide wider bandwidth and quicker process ability. Packet classification is essential for policy-based firewall deployment, intrusion detection, differentiated services, and Virtual Private Network. At the same time, with the wide use of high-speed IP network, most of implements of network packet capture cannot be adapt to the need of traffic monitoring on the high-speed network, excessive CPU utilization and serious packet loss have become a urgent problem to be solved for the traditional implementation. So the research in packet classification is significant.Firstly, the paper briefly narrates the packet capture technology itself and its development of a domestic; then presents and analyzes on the structure of the BPF (Berkeley Packet Filter) Packet filtering mechanism widely used on Linux and the traditional implementation of application layer packet classifier by string match , and studies on the main factors which cause the low performance result for traditional implementation; then the paper provide a scheme to classify network packet in high efficiency by combination of hardware and software. The scheme firstly preparatory filter the packet by hardware, then the software part classified the packet based on the result of hardware filter, to provide high efficiency packet classifier.Secondly, the paper describes a design and implementation of a SIP Signaling capture platform for 2.5G bit network base on above theoretical research. The paper also tests the traditional capture platform and the improved capture platform respectively on a simulated 2.5G bit network, and analyzes the result. And find out that the new scheme is better then traditional scheme both in packet losing ratio and resource consuming.Finally, we draw a conclusion about the paper and give some advice on the improvement of the system further.