基于虚拟资源的资源调度算法研究

发布时间：2018-03-14 06:03

本文选题：云计算　切入点：资源调度　出处：《中国科学技术大学》2014年硕士论文　论文类型：学位论文

【摘要】：随着“云计算”的出现和快速发展,云计算正逐渐从理论走向实践,“云”作为一种新型的资源形式被越来越多的用户所使用,云计算也成为学术界和产业界研究的重点。云计算的关键是利用现在成熟的虚拟化技术,将传统的基础设施虚拟化成一个公共计算资源池,然后对资源池中的资源统一调度分配。因此云环境中的资源调度问题成为了云计算中不可忽略的问题。同时云提供商为满足用户的不同需求,提供了不同种类的虚拟资源。而当前主流的资源调度算法主要是从系统节能或者请求服务质量等角度考虑问题,并没有考虑到云提供商提供的虚拟资源之间的差异性。本文作者在调研云计算关键技术,特别是虚拟化技术的基础上,调研和总结了目前云计算中常用的资源调度算法,并在前人算法的基础上,考虑了虚拟资源间的差异性,提出了基于虚拟资源异构性的资源分配算法RA-HR;同时为确保系统较低的资源使用率和虚拟资源间的公平性,作者提出了双层的资源调度框架DTRAC。 RA-HR算法首先从用户任务包的全局角度考虑,将用户提交的云任务按照最少剩余内存原则划分成不同的组合,再根据虚拟资源间的差异性,为相应的组合分配相应的虚拟资源。DTRAC包括请求调度层和局部资源分配层两层。请求调度层利用反馈回来的服务器资源使用情况,将应用请求调度到资源使用率最低的服务器上；资源调度层首先将服务器上虚拟资源分层聚类,然后按照应用请求的需求比例分别从各个类中为请求分配资源,以确保同一个服务器上各个虚拟机之间分配资源的公平性。文中通过Cloudsim工具对算法进行了仿真。在仿真RA-HR时,文中通过对虚拟资源充分和不充分两方面进行仿真看出：在云提供商提供不同类型虚拟资源时,RA-HR在虚拟资源的使用情况上优于传统算法,表明RA-HR算法能在满足用户需求的前提下,在节省虚拟资源使用上有较好的表现；同时通过对DTRAC的仿真得出：DTRAC在系统资源使用率、分配资源公平性上优于传统算法,表明DTRAC可以有效减少系统资源的资源使用率,在分配虚拟资源的公平性上有不错效果。
[Abstract]:With the emergence and rapid development of cloud computing, cloud computing is gradually moving from theory to practice. As a new type of resource, cloud is used by more and more users. Cloud computing has also become the focus of research in academia and industry. The key to cloud computing is to virtualize traditional infrastructure into a pool of public computing resources using today's mature virtualization technologies. Then the resources in the resource pool are allocated uniformly. Therefore, the resource scheduling problem in the cloud environment has become a problem that can not be ignored in cloud computing. At the same time, in order to meet the different needs of users, the cloud providers, The current mainstream resource scheduling algorithms are mainly from the point of view of system energy saving or request quality of service, and do not take into account the differences between the virtual resources provided by cloud providers. On the basis of investigating the key technologies of cloud computing, especially virtualization, the author investigates and summarizes the resource scheduling algorithms commonly used in cloud computing, and considers the differences between virtual resources on the basis of previous algorithms. A resource allocation algorithm RA-HR-based on the basis of the heterogeneity of virtual resources is proposed, and in order to ensure the low utilization rate of resources and the fairness between virtual resources, a new resource allocation algorithm, RA-HR-based algorithm, is proposed. In this paper, a two-tier resource scheduling framework, DTRAC. RA-HR algorithm, is proposed. Firstly, the cloud tasks submitted by users are divided into different combinations according to the principle of minimum residual memory, and then according to the differences between virtual resources, the proposed algorithm is based on the global perspective of user task packets. The corresponding virtual resource. DTRAC includes two layers: the request scheduling layer and the local resource allocation layer. The request scheduling layer schedules the application request to the server with the lowest resource utilization rate by utilizing the feedback server resource usage. The resource scheduling layer firstly clusters the virtual resources on the server and then assigns the resources from each class according to the requirement ratio of the application request to ensure the fairness of the resource allocation among the virtual machines on the same server. In this paper, the algorithm is simulated by Cloudsim tool. In the simulation of RA-HR, Through the simulation of virtual resources, it is shown that when cloud providers provide different types of virtual resources, RA-HR is superior to traditional algorithms in the use of virtual resources. It shows that the RA-HR algorithm can save virtual resources on the premise of satisfying the needs of users, at the same time, through the simulation of DTRAC, it can be concluded that the utilization rate of system resources and the fairness of resource allocation are better than those of traditional algorithms. The results show that DTRAC can effectively reduce the resource utilization rate of system resources, and has a good effect on the fairness of allocation of virtual resources.
【学位授予单位】：中国科学技术大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP393.01

【参考文献】