A Cloud Service System Based On Memory Template Of Virtual Machine

In data centers and cloud computing environments, the number of virtual machines (VM) increases with the services'requests increased. When these services are switching and scheduling constantly with the requests, the corresponding virtual machines will also be created on and off frequently. These factors cause the time of creating a virtual machine to be the bottleneck of data centers services. Besides, the virtual machines need to read image file from the disk to start themselves, which cause a great loss for performance of the disk I/O.In this paper, we present a cloud service cache system based on memory template of virtual machine to improve the response time of the cloud computing service and reduce performance loss of the disk I/O. This system can create and maintain service cache VM through the memory template, which is a snapshot of a running virtual machine. Then, the subsequent service can be deployed directly on the virtual machine rapid deployed, avoiding the waiting time and performance loss of disk I/O brought by creating virtual machines from image file.The cloud service cache system is implemented based on Xen hypervisor. We hack the origin code of Xen tools and add the new feature for creating VM with memory template, which reduces the time spent on creating VM. To evaluate our system, we focus on demanding of search requests based on Hadoop. The experimental results show that the average time of creating VM reduces about 80% and the amount of disk access data decreases above 50%, the time spent on WordCount of Hadoop decreases 40%.