The Multi-objective Optimization Algorithm In Cloud Task Scheduling Based On The Improved GEP And ANP

Cloud computing is consisted of a series of scalable resource pools, through which the consumers can rend the resource by dynamic virtualization technology. Because of the expanding number of cloud consumers and the increasing size of the cloud data center, how to make better use of the network resources has become one of cloud computing problems which is urgent to be solved. The main works in this paper are as follows:(1)We proposed the mathematical model of the cloud task scheduling based on the matrix theory, which contains the maping matrix, cost matrix and transition matrix. It provides the theoretical basis and mathematical support.(2) Aiming at the problem of the existing cloud scheduling algorithms which treat the cloud task scheduling problems on optimization goals and scheduling level in a relative single way. A dynamic task scheduling model with multi-objects and multi-scheduling levels has been proposed by introducing the NSGA-II into the field of cloud task scheduling.(3) Aiming at the deficiencies of the traditional multi-objective optimization algorithm which shows the limited optimization abilities, a novel multi-objective optimization algorithm based on NSGA-II and improved GEP has been proposed by combining the two algorithms. Some modifications about corresponding mode and operation of GEP have been made to adapt the GEP to the cloud task scheduling model and NSGA-II proposed in this paper.(4) The ANP model has been added into the task scheduling strategy to make the Pareto solutions more readable and help the consumers make the most suitable decision among the Pareto solutions which contains all the possibilities.In summary, a dynamic multi-objective multi-task scheduling model has been proposed, and the mathematical description of the cloud task has been ameliorated. At the same time the improved GEP and ANP model have been added in the NSGA-II to get the global optimization abilities of the NSGA-II improved. A series of simulation experiments have been made, whose results show that the algorithm proposed in this paper is feasible and effective.