The Simulation And Research Of Task Scheduling In Grid Environment

Grid technology links all kinds of resources shared in the Internet, such as high performance equipment, expensive instrument, storage equipment, software and database, etc, and converts them to an available, reliable, standard and economical computing power. In order to process tasks on proper resource, grid schedule system need match and schedule the submitted tasks to resources.Task matching and scheduling plays an important role in network computing and makes a notable impact on the overall performance. Scheduling problems are known to be in general NP-complete, only sub-optimal can be obtained by classical scheduling approaches in most cases.This paper studied task scheduling algorithms using improved genetic algorithms for network computing by means of theoretical analysis and simulation experiments. Genetic algorithm is a kind of random searching method directed by fitness function. It simulates the biology evolution in nature in order to find the best value.We first proposed a general genetic algorithm for task matching and scheduling in grid environment. A chromosome is specified by a resource-task matching matrix. Three genetic operators, such as external crossover, internal crossover and migration as a kind of mutation, were designed based on permutation representation. The algorithm can avoid the premature phenomena and break out of the trap of the part best value using three operators and the height of task.A large mount of experiments had been performed to obtain a range of main control parameters settings for proposed algorithm. The DAG of tasks, the number of tasks, the MI of the tasks, resources, execute efficiency (MIPS), delay can all be set automatically. The emulational experiment results show that the genetic algorithm proposed in this paper has faster convergence speed and greater probability to find optimal value.