Research On Task Scheduling And Virtual Machine Migration Strategy In Cloud Computing

Task scheduling and virtual machine migration are important issues of resource scheduling and management in the cloud computing environment.In this paper, third and fourth chapters focus on the task scheduling strategy based on improved genetic algorithm, studying how to dispatch tasks to the most suitable virtual machine to carry out and implementing the shortest completion time of tasks and the maximum utilization rate of the virtual machine. The fifth chapter mainly study virtual machine migration strategy based on the energy consumption awareness, through the migration of virtual machine on low load host to another host and close the low load’s host to achieve the purpose of energy saving.On the basis of summarizing the predecessors’ research work, the paper analyzes the existing problems and gives the corresponding solution.This article mainly has the following four working content.This paper firstly introduced the research status quo of task scheduling in cloud computing environment at home and abroad, and it expounded the concept, classification, key technology and Map/Reduce distributed programming model in detail, and analyzes the two levels of resource scheduling model in cloud computing, the target and characteristics of task scheduling and common algorithms for task scheduling.For traditional genetic algorithm exists slow convergence speed, easy to premature and poor local search ability shortcomings, this paper proposes a single-oriented population based on hybrid genetic algorithm and simulated annealing algorithm of task scheduling strategy. The algorithm’s main innovation point lies in the integration of the thought of simulated annealing algorithm, accepting inferior solution at a certain probability in the process of the optimization, and increase the diversity of individuals in the population evolution structure, providing power for algorithm late iterative optimization, reducing the possibility of algorithm falls into local optimum.The improved algorithm main approach is to optimize the function of individual choice and give an adaptive crossover and mutation probability functions, to enable them to correctly guide the direction of evolution of the population.The paper based on consideration of the timely response to the user and supplier interests,we designed the dual fitness function based on average task completion time and load balance.The simulation experiments show that compared to Round-Robin, genetic algorithm and simulated annealing algorithm, the proposed algorithm has shorter completion time and larger virtual machine resources utilization.Aiming at the shortcomings of the single population evolution, the fourth chapter puts forward the multiple population annealing genetic algorithm to improve cloud task scheduling. Different from single species, multiple population changed the population of a single way of evolution.It adopt gene exchange between the populations and excellent individual migration strategy, enriching the population evolution way and increasing the diversity of the individuals in the evolutionary process. Simulation results show that the proposed algorithm than third chapter’s algorithm has shorter task completion time, and can more quickly find the optimal scheduling results.Based on the secondary resource scheduling model in cloud computing, the fifth chapter studied the virtual machine migration issues in the resource allocation and put forward a kind of virtual machine migration strategy based on energy consumption awareness. The main idea is to migrate virtual machine on the server whose resources utilization is below a certain threshold, by closing or dormancy the server that is not running any virtual machine, to achieve the goal of data center energy saving.