Research And Implementation Of The Virtual Machine Scheduling On Cloud Platform

Cloud computing is developed and fromed by the grid computing, distributed computing, parallel computing, breaked the traditional IT service model. Through the virtualization technology to make physical resources virtualization, then form the vitual resource pools. The data center to manage it uniformity, allocate it dynamically and provie transparent service to users. Virtual machine as a resource allocation to users, the different mapping between the virtual machine and the physical machine brings different power consumption, resource utilization and so on. Therefore, the virtual machine scheduling problem is an important issue of resource management in cloud computing data center. In recent years, there have been some research results about the virtual machine scheduling problem, but most of the researches are focused on the server.This thesis describes and analyzes the current research situation of virtual machine scheduling, summarize and discover: on the one hand, there is little research on the scheduling problem of the related virtual machine group, but ignoring the relationship between the virtual machine can cause the users’ application performance. On the other hand, facing the multi-objective optimization scheduling problem is current research hotspot, related to the selection of multiple optimization targets and the application of the optimization algorithm. At present, the research on the selection of multiple optimization objectives is less about the user service quality. In view of the above two questions, the main research contents of this thesis are as follows:(1) The problem of associated virtual machine scheduling is studied. In the study, to reduce the total network traffic and improve the communication efficiency, the virtual machine with large network traffic is placed together. At the same time, in order to avoid excessive concentration of link congestion, the maximum link utilization is proposed to minimize the network traffic distribution, and reduce the congestion rate. Therefore, The scheduling method involves two indicators: network traffic and link utilization. Using improved ant colony algorithm to solve this problem. In the experiment, VMPACS and PSO are selected as the contrast algorithm, compared to this two algorithms, the improved algorithm in this thesis, respectively, decreased by 14.1% and 12.8% in the reduction of network traffic. compared to a single target to optimize network traffic, the maximum link utilization of HACO_SA algorithm has been significantly reduced.(2) Multi objective optimization of virtual machine scheduling problem is studied. To the cloud data center power consuming huge this problem as the breakthrough point, this thesis puts forward the balanced utilization of each dimension of the server resources, and avoids the waste of resources caused by the barrel effect, at the same time can not ignore the SLA violation rate in the optimization of the above two goals. This is a classical multi-objective optimization problem, Using the grouping genetic algorithm to solve this problem, the fuzzy logic method is used to solve the problem of the uncertain weight in the multi-objective optimization problem. Compared with the GGA algorithm and MMA algorithm, the performance of the algorithm in this thesis is improved by 11.4% and 5.8% respectively in the multi objective virtual machine scheduling problem.(3) Design and develop a virtual machine scheduling system platform based on OpenStack, which verifies the validity and rationality of the proposed scheduling algorithm.