Research On The Network Load Balancing Algorithm Based On Simulated Annealing Genetic Algorithm

With rapid development and widely usage of network technology, the online information exchange amount increased sharply, customer's requirement in performance and security of applications has become more and more rigorous. Traditional hardware's upgrading is unable to satisfy with the demand of application. By using several servers to share load, we can use these servers to deal with the same service or use different servers to handle different services. But with the network paroxysmal, it could be difficult to confirm which pages make the large load. The basic solution to deal with the large load network is using the technology of network load balancing.Based on the purpose of the network load balancing, this paper combined Genetic Algorithm with Simulated Annealing Algorithm appropriately, formed a new algorithm that is fit for network load balancing. This algorithm improve the performance of the traditional Genetic Algorithm, accelerate the convergent speed of the algorithm, strengthen the whole astringency, and also increase the reliability of finding the whole optimal solution. With this new algorithm we can make the full use of the network sources and enhanced the capability of the network load balancing. The main research of this paper is as follows:1, We design a new coding mode of number sequence based on the thought of the load balancing. This coding mode is simpler and more intuitionists and more fit the need of the real network. There is no need of decoding while evaluated the individuals. What's more we put forward a mend of the number sequence so as to avoid break the rules of priority restrict.2, After analyzed the deficiency of Genetic Algorithm, we improved the fitness function and its operations of select, crossover and mutation in traditional Genetic Algorithm to fulfill the purpose of the network load balancing which is to reduce the response time and to increase the usage of the resources. Still we introduce the Simulated Annealing Algorithm to the fitness function and add the annealing temperature in it to get the following result: The individuals with close fitness degreehave the similar probability of their offspring when the temperature is high (the earlier stage of GA). While the temperature declined constantly, the fitness function shows its power and magnified the difference among those close individuals, so as to make the advantages of the outstanding individual more clearly.3 -. Instead of using the roulette selecting method, we choose a selecting method which choosing probability is pro rata with the value of the fitness, and add an optimal saving technique. With this method the individual with good fitness can be saved to the next generation, thus accelerate the constringency of the algorithm effectively.4^ By studied the method of the crossover probability Pc and mutant probability Pm automatic changed with the value of fitness function which is brought forward by Srinivas. In order to accelerate the constringency of the algorithm we established a mode of Pc and Pm to make the individual with a small fitness value have a larger crossover probability and mutant probability. When the algorithm traps in partial extremism the Pc and Pm of the individual with large value will increase, so as to avoid the matter of "early-maturing".