Improvement Of Cooperative Co-Evolutionary Algorithm And Its Application In The Tasks Scheduling Of Cloud Computing

At present, Genetic Algorithm attracts the attention of researchers with its unique advantages. But the coding of genetic algorithm is too long when solution space is large, it is not convenient to handle. So the cooperative co-evolutionary algorithm is born. It inherits the advantages of genetic algorithm, and overcome the shortcomings of it. So the algorithm is put forward from the date of birth.At present, cloud computing technology is developing rapidly. Cloud platform need to deal with huge amounts of user requests. How to reasonably scheduling user’s tasks to meet the needs of users, is an urgent problem in the development of cloud computing technology.Because there are advantages when the cooperative co-evolutionary is adopted to deal with the complex multivariable problems,it is current’s research hotspot to use the cooperative co-evolutionary algorithm in the cloud task scheduling problem.An individual of each population algorithm represent only part of the solution in cooperative co-evolutionary, so we need to choose collaborators from other populations to form a complete solution to evaluate the advantages and disadvantages of the individual. Collaborators selection problem is a very important problem in cooperative co-evolutionary algorithm. Currently, there is no suitable solution for collaborator selection problem. There is room for improvement in this area. This paper presents a new collaborator selection method based on distance according to the classification in machine learning. The algorithm calculated the distance of the individual to be evaluated to the optimal individual and the random individual to select the most appropriate cooperative groups. This method can make a more reasonable evaluation of the individual as well as control the counts of evaluation. Thus, the algorithm can get a better solution. The availability and validity of this algorithm are verified by experiments on the typical function optimization problem as well as on the Job Shop scheduling problem. The experiment proved that the algorithm can get a better solution.This paper adopted improved collaborative evolutionary algorithm to solve cloud task scheduling problem, the main purpose is to solve the problem of huge user tasks request quantity and time span in the cloud task scheduling. First of all, the cloud task scheduling is abstracted to an optimal model; Then designing the coding method and genetic operators’ operating details, so the algorithm can play the best performance; Finally, designing the overall scheduling process of using improved cooperative co-evolutionary algorithm to solve the cloud task scheduling problem. The experiments are conducted on the simulator Cloud Sim. Experiments show that the algorithm can get better time span than mainstream scheduling algorithm when the performance of virtual machine in the data center is almost the same. The algorithm can get better time span than GA(Genetic Algorithm) and CCGA(Collaborative Co-evolutionary Genetic Algorithm) when the performance of virtual machine in the data center differs greatly, but is bad then MIN-MIN algorithm. As a result, the algorithm is not suitable for cloud task scheduling problem when the performance of virtual machine in the data center differs greatly.