Research On Models, Algorithms And Implementations Of Traffic Control Based On Network Processor

The rapid evolution of Internet requires high-speed and flexible network traffic control devices, but traditional ASIC based or software based schemes can not meet the both requirements at the same time. In this thesis, the two key issues of high-speed traffic control (i.e. QoS control and security control) are studied based on the emerging Network Processor (NP). Models and algorithms are not only studied in theory, but also implemented based on NP. The followings are the detail contents and contributions of the thesis.Firstly, an integrated queue management algorithm fitting for NP is pre- sented. Our algorithm which avoids division and sort operation, preserves O(1) time complexity and can assure the proportional service in both loss ratio and delay. The experiments on Intel IXP2400 show that our algorithm can achieve the proportional differentiated service and reach 1Gbps wire-speed.After that, a multidimensional classification algorithm called AM-Trie (Asymmetrical Multi-bit Trie) is presented. AM-Trie is a high-speed, parallel and scalable algorithm and very fit for the"multi-thread and multi-core"property of NP. An optimized field division algorithm is also presented and proved. Our algorithm can reach 10Gbps wire-speed in IXP2800 platform.After studying the features of future Internet and worm propagation, I present the WSMN (Worm Spread on Mobile Network) model and find the critical condition of worm fast die out. Extensive experiments have been performed to verify our WSMN model and the critical condition.Based on the analysis of WSMN model, an anti-worm algorithm is designed based on parallel Bloom filter which makes the most of the parallel processing property of NP. The influence of each parameter also discussed. The anti-worm filter can reach 500Mbps in IXP2400 and the experiments in real environment against Worm Blaster prove its effectiveness.