A Fast Approach of Provisioning Virtual Machines by Using Image Content Similarity in Cloud

Abstract

To quickly provision multiple virtual machines (VMs) is a challenge in nowadays cloud data centers (CDCs). By utilizing the content similarity among the virtual machine image (VMI) files, the amount of data transferred in the VM provisioning is reduced, and hence, the provisioning time can be shortened. Thus, minimizing the total amount of transferred VMI file data is helpful for accelerating the VM provisioning. Meanwhile, packing the VMs into the minimum number of physical machines (PMs) is also crucial for the CDCs. To solve these two problems at the same time, we propose a heuristic algorithm, called fast balance placement (FBP), by utilizing several tables to precompute and store the similarity relationships among different VMI files. Comparing to the balance-placement algorithm, the simulation results show that FBP uses less PMs to pack the VMs and its running time is shorter, and it transfers almost the same amount of the VMI file data.