Resource Scheduling In Cloud Computing Based On Co-evolutionary Genetic Algorithm

As a provider of dynamic services using very large scalable and virtualized resources over the Internet,cloud computing receives wide attention and research in academia and industry. Cloud computing platform integrates various resources into a virtual resource pool from which users could access to computing resources, storage resources and software services according to their different needs. How to pool and management resources for resource scheduling, and the scheduling efficiency directly affect working performance of the entire cloud computing platform. Besides, cloud computing provides services for a lot of users who have a variety of targets. These come out a variety of resource scheduling algorithms in cloud computing platforms. In this thesis, cloud computing architecture and business models are described, and research status of scheduling algorithm in cloud computing are conducted a comprehensive analysis. Aiming at some shortcomings of the existing scheduling algorithm, the main research work is as follows:(1)An improved cooperative co-evolutionary genetic algorithm is proposed for solving multi-objective scheduling problems in cloud computing environments. When users perform tasks using cloud computing platform, they need to hire virtual machine resources from cloud computing platform, which results in execution costs. However, the users want to complete the tasks as quickly as possible. As the execution cost and completion time are two goals conflicting with each other, users can only obtain the equilibrium outcome of them. In the cloud computing research, people usually put the two objects into a single object by weighting method, each can only provide a scheduling scheme for the users, and many studies have neglected the dependencies between tasks. Therefore, this thesis, by modeling the task scheduling with precedence constraints, Propose a multi-objective problem having execution cost object and completion time object. For solving this multi-objective problem, an improved cooperative co-evolutionary genetic algorithm which uses the Taguchi method for exploiting the better individuals on micro-space and multi-population with the outside optimal solution for exploiting solution in the global search space is proposed.(2)In previous studies about cloud computing scheduling algorithm, people usually ignore impact of data transmission between tasks, but data locality and data transmission have a great influence on the data-intensive task, especially for scientific workflow scheduling. This thesis explains the importance of data locality and data transmission.Taking into account that the users are more concerned about the completion time, a competitive co-evolutionary immune genetic algorithm, combining with immune algorithm and competitive co-evolutionary algorithm, is proposed for scheduling problem in scientific workflow for data intensive tasks, which aims to minimize the completion time.(3)In the cloud simulation platform building by Cloud Sim, the two scheduling model and the corresponding algorithm proposed in this thesis are carried out experiments. The results demonstrate the effectiveness of the two proposed scheduling algorithm.