Grid Resource Scheduling Strategy Based On Improved Genetic Algorithm

Grid is an environment that integrates calculation and resources. It can fully absorb all kinds of computing resources and turns them into a computing capability which is easily available,reliable,standardized and economical.In this way,we can share the resources thoroughly.Task scheduling is one of the most important part in grid research,How allocate tasks to different resources in order to make the grid system obtain the highest performance is the problem which the task scheduling needed to resolve.The features of distribution,heterogeneousness,dynamic and self-ruling of grid challenge the traditional scheduling algorithms. So it is very important and realistic to put forward a better scheduling algorithm based on existing algorithms which can make full use of all kinds of resources and improve the throughput of grid system.Genetic algorithm is a new kind of modern heuristics algorithm,which is often used to solve different kinds of NP-complete problems and complex scheduling problems. Some simulation experiments have confirmed that genetic algorithm is better than classical scheduling algorithms.Because the genetic algorithm (GA)has some defects,such as prematurely convergence and deceptive problem, many academicians committed themselves to the research of improving the genetic algorithm.Traditional adaptive genetic algorithm(AGA) has higher convergence speed,but it still easily gets stuck at a local optimum.This thesis analyzes the basic principle of the genetic algorithm(GA) and cloud model theory, Aiming at the shortage of GA and AGA, this thesis brings forward a kind of a cloud model adaptive genetic algorithm(CMAGA).Some improvent is made on the basis of GA and AGA:selection operation uses the elite strategy and retain the optimal selection ,so we can enhance the diversity of the species. CMAGA is based on cloud model with the propertics of randomness and stable tendency.In the CMAGA,the probabilities of crossover and mutation are adaptively varied depending on cloud generator . CMAGA can improve its convergence capacity because of the stable tendency of cloud model. Meanwhile, it can remarkable avoid a local minimum using the randomness of cloud model to maintain diversity in the population.Based on characteristics of task scheduling in Grid, we design all the components of CMAGA, at length.On top of that,we perform an experiment based on Matlab language to implement the algorithm,and do comparison with GA and AGA in performance. The experiment results show that the reformative genetic algorithm not only has a holistic searching ability, but also make a fast convergent speed, which provide a preferable performance. This experiment realizes the optimal makespan of the task scheduling.