Research Of Virtual Machine Scheduling Method For Campus Cloud Platform

The rapid development of information technology and the continuous deepening of education reform promote the construction of college information. Since a data center is mostly based on the servers and SAN storage during the process of building, an application is deployed on one or more servers, resulting in a low utilization of the infrastructure, laboratory equipment islands and so on. The rising of cloud computing technology provides a feasible scheme for the problems above-campus cloud.The campus cloud mainly uses the virtualization technology to integrate and virtual physical resources to achieve the purpose of dynamic allocation and on-demand use of resources. However, the application scenarios of campus cloud is closely combined with teaching activities, the characteristic of teaching application is periodicity, predictability, and batch, unlikely with the traditional application. The current virtual machine scheduling mechanism doesn’t take into account the characteristics of campus applications, resulting in a lot of unnecessary migration of virtual machines, the waste of resources and load imbalancing. This paper aims to solve the problems, proposes a virtual machine scheduling method and system for the characteristics of teaching application of campus cloud platform, the specific research work and the main contributions are as follows:Propose the theoretical model of the scheduling problem for teaching application(1)First, abstract the teaching application scheduling problem, use the formula to define the purpose of making the number of active physical machine as little as possible, each active physical resource utilization as balanced as possible; then proposed the course requirement model and physical machine load model. Course requirement model is defined as a six-tuple, the model is the abstraction of teaching application load, solving the problem of the traditional model can’t effectively use the course information to judge the decision. Physical machine load model is defined as a three-dimensional matrix of a5*5*16to describe the details of the course VMs deployed on a physical machine. It is carried out in the traditional model, can describe the using time of virtual machine and a maximum resource utilization for a set time, helping make more precise scheduling.(2)Propose scheduling strategies and algorithms for teaching application in campus cloud platformTo solve the problem of uneven load and waste of resources, this paper present the scheduling strategy for periodicity, predictability and batch of teaching application, proposed deployment principle of "first set, and then disperse" and dynamic scheduling theory of "first balance, after saving", defined goodness of fit, deployment priority value, migration priority value and load decision value algorithm, and ultimately proposed course requirement deployment algorithm and dynamic scheduling algorithm.(3)Propose scheduling system architecture of campus cloud platform and realize the systemBased on the scheduling method for the campus cloud platform proposed in this paper, the paper designs the of the scheduling system architecture. The main module of the system includes user request module, data storage module, the virtual machine scheduling module and the virtual machine control module. Finally the scheduling system is implemented on the campus cloud platform based on OpenStack.Experiments show that the proposed scheduling method can guarantee the need of teaching application, and effectively reduce the number of active physical servers, realize load balancing at the same time, achieving the purpose of multi-objective optimization.