Implementation And Improvement Of Task Scheduling In A Distribute Computing System Based On Genetic Algorithms

Due to the difference between the characteristics of different tasks and the performance of the computer, job scheduling has become the key factor that affects the efficiency of distribute computing systems.Job scheduling problem has been proved as a NP-complete problem,which means that exhaustive search is the only way to achieve the optimum solution. How to approach the optimal solution in a short time by the reasonable algorithm is the main problem of the research on the distributed task scheduling, also the main problem of the research of this thesis. The main work is as follows:1. Aiming at the premature problem of genetic algorithm, a genetic algorithm based on geographical isolation is proposed. This algorithm can suppress premature phenomena, prevent the algorithm from falling into local optimal solution, and improve the computational efficiency of the algorithm, reduce the convergence time of the algorithm.2. Aiming at the problem that the fitness values of random initial population of genetic algorithm are too low, a self optimization algorithm is proposed. This algorithm can quickly improve the fitness values of the initial population, so as to reduce the overall time of computation3. The above mentioned algorithms are combined and applied to the task scheduling module. Then, a distributed computing system is designed and implemented. Compared to other implementations,this system can provide faster and more efficient task allocation scheduling,greatly save the computing resources and reduce the load of the central server, and can also enhance the throughput of the system.