Fast Provisioning of Virtual Machine Based on Intelligent Virtual Machine Image Prefetching.

Virtual Machine (VM) provisioning is one of the most critical issues in cloud computing. However, the long delay in VM provisioning significantly degrades the customer experience. Especially, the time-consumption of transferring Virtual Machine Images (VMI) from storage nodes to selected compute nodes is one of the major factors that cause the delay in VM provisioning. Unfortunately, existing approaches usually make their efforts to improve efficiency at the cost of disk performance. In this paper, we propose an intelligent VMI prefetching approach to expedite VM provisioning. First, we identify the data blocks necessary for VMs startup from the whole VMI and record their access order. Then, we design a VMI prefetching mechanism to reduce data transmission, which transfers only a small amount of startup block to compute node during the pre-steps stage (i.e., the short time before startup). Next, we employ a VM request prediction model which leverages historical records to predict the number of requests for each VM type in the future period, so as to perform VMI prefetching in advance, thus reducing VM suspension caused by the preparation of requested blocks. Furthermore, we implement a prototype named PreVMI and perform a series of experiments to evaluate the performance of our approach. Experiments show that our approach can reduce VM provisioning latency by about 35%.