The Research And Optimization Of Network Traffic Monitoring Based On Multi-Core Platform

Network traffic monitoring is a necessary requirement when the Internet development in a certain stage. One can better understand and monitoring the Internet, it can also provide a better service. With the growing of network bandwidth and the emergence of new applications, traffic monitoring system is facing unprecedented pressure. Everyone hopes that the traffic monitoring system can work fast and give accurate results. This makes a high performance demands for traffic monitoring system. When facing huge network traffic that beyond the processing capacity, traffic monitoring system will appear many performance problems such as losing protocol status, missing key information and abnormal program termination, which cause traffic monitoring system to be not available. People want to find new solutions to improve the processing performance of traffic monitoring systems.General-purpose multi-core processor has strong parallel computing capability. It provides high-speed concurrent computing engines by integrating multiple cores in one chip and maintaining or reducing overall energy consumption. For multi-core processor, it needs compute-intensive applications to avoid the waste of computing resources. Multi-core processor not only provides assurance of computing, but also makes researchers to see the direction of performance optimization.This thesis is the intersection of two aspects above, including some knowledge of network traffic monitoring, computer architecture, computer operating systems, etc. We do performance optimization of traffic monitoring system on the multi-core platform by combining traffic monitoring technology and multi-core technology. In this thesis, we mainly do the researches as follows:We study the parallel optimization technology of software on the multi-core platform. It includes parallel computing, processor instructions, multi-core performance optimization and so on. Furthermore, by reading a large number of references, we summarize eight methods of performance optimization on multi-core platform. These methods can be applied in different kinds of software to improve their performance. We think these methods are valuable reference for performance optimization of traffic monitoring system on multi-core platform.We proposed a general architecture of traffic monitoring system and a set of evaluation indicators which are used for optimizing system performance on multi-core platform. In this thesis, we summarized the general architecture of traffic monitoring system in order to show the system function. We also proposed a set of evaluation indicators for traffic monitoring system which used for improving performance on multi-core platform. This set of evaluation indicators includes key indicator and assistant indicators. The key indicator is the system throughput. The assistant indicators could be further classified into three categories which are used to evaluate system overhead, scheduling strategies and the optimization effect. These evaluation indicators can help us improving system performance on multi-core platform.We took a specific system which belongs to the category of Network Protocol Analysis as a study case to improve its performance. On this basis, we summarize the methods and features of performance improvement which is suit for this category of system on multi-core platform. The specific system is named'GTP Analysis System'. We analyze basic performance of the system, find performance bottleneck, and finally propose optimization strategies on multi-core platform. The experimental results show that when using seven cores to compute, system throughput is improved to 401.73% compared with before. Then we summarize the performance optimization features of Network Protocol Analysis System on multi-core platform.After studying the optimization features of Network Protocol Analysis System, we chose another category of system called Network Traffic Content Monitoring System. We took a specific system which belongs to this category as a research case to improve its performance. On this basis, we summarize the methods and features of performance improvement which is suit for this category of system on multi-core platform. The specific system is named ITCMS (Internet Traffic Content Monitoring System). Similarly, we evaluate the basic performance of ITCMS and analyze performance bottleneck. Then we proposed optimization strategies. Experimental results show that when using seven cores to do the optimization, system throughput is improved to 436.10% than before. Next, we summarize the performance improvement features of Network Traffic Content Monitoring System on multi-core platform. Furthermore, we do a comparative research on optimization features between these two categories of system.In the last part of this thesis, we analyze performance of packet capture on multi-core platform, and we also do a research on optimization of packet reassembly system based on multi-core platform. For packet capture performance, we chose two common approaches, namely PF_PACKET Socket capture and Libpcap capture. The analysis contains two aspects:One is the maximum packet capture rate and throughput on multi-core platform. The other is the CPU load under the maximum capture rate. In both of the two aspects, we explore the performance of single capture process and dual capture processes separately. For packet reassembly optimization, we chose Libnids as a study case. We analyze the system performance bottleneck and parallel the reassembly system according to the HTTP protocol features. When we use two cores to compute, this optimization improves the throughput of packets reassembly system to 145.37% than before. The performance result can meet the basic requirements of common network uses.