关键词: 包匹配, 差分演化算法, 变异系数

Abstract: The development of Internet has transferred the bottleneck of network speed from link speed to packet processing speed of the core network equipment, the critical work of packet processing is packet matching. The speed of packet matching is difficult to adapt for packets linear forwarding of the core network equipment by traditional algorithms. This paper proposes a novel algorithm of packet matching by combining improved differential evolutionary algorithm and classic packet matching algorithm. To increase objectivity, the method of statistics is used in computing the fitting value. Experimental results show that this new algorithm effectively improves the performance at the speed, storage space and update time, compared with the traditional one, and also show that the time performance of the proposed algorithm in packet matching is very weakly correlated with the number of rules, thus the algorithm is very suitable for solving multi-field and large-scale problems. The paper uses evolutionary algorithm to solve the network data packet forwarding in the condition of multi-field and large-scale rules warehouse, and packets can forward in the linear speed. This new algorithm is universal and adapts for many equipments such as firewall and QoS router etc.

Key words: packet matching, differential evolutionary algorithm, coefficient of variation