Research And Implementation Of The Technology Of Virtual Machine Grouping Connection Pool Based On Cloud Computing

With traditional allocation algorithms, resources of virtual machines have to be built, allocated and recycled repeatedly according to user’s requests in a cloud computing environment. It makes allocations costly and decreases the performance of a cloud operating system. As with this problem, this thesis proposes a novel grouping connection approach for resource pools of virtual machines, which improves allocation strategies of the traditional virtual machines. It groups a large number of virtual machines by creating templates, each group of which corresponds to different requirements of cloud tenants. When the system begins to initialize, virtual machines of appropriate configuration have been created in each group, which decreases cost and improves performance. On the basis of the above work, this thesis proposes a system framework, which implements the grouping connection approach for resource pools of virtual machines. Specifically, this thesis made the following research achievements:(1) As with the problem of dynamic management strategies for the connection of virtual machine and optimal fair allocation of grouping resources of virtual machines, an algorithm is proposed to postpone the release to solve the problem of virtual machine resource reuse. This algorithm can dynamically decide whether the connection resource of virtual machine is released or not according to the number of requests and minimize the connection time for creating new virtual machines. It handles reusing resources of virtual machines, hence greatly improving the reuse rate of virtual machine resources and effectively reducing the overhead to create and recover resources of virtual machine resources.(2) Based on an open source cloud platform CloudStack, we designed and implemented the system of virtual machine grouping connection pools, including the following functionalities: authority control, resource creation, resource monitoring and resource scheduling. Additionally, we also designed a system database and provided a solution to centralized exception handling.(3) Experimental results show that the system of virtual machine grouping connection pools satisfies the requirements for performance. Compared with the traditional allocation algorithms, our algorithm decreases the user response time of dynamic virtual machine grouping connection pools significantly by roughly 50% in the cloud environment. It enhances the performance of the whole cloud operating system and improves user experiences.