A Performance Mapping Model for Physical-to-Virtual Migration

Abstract

With the increasing popularity of cloud computing, hosting services on a remote virtual machine becomes a trend due to its ease of use and cost saving. When migrating services to a cloud, users need to submit a resource plan to ensure service's performance. However, it is difficult to estimate the resources required by a virtual machine for running arbitrary services. To address the problem, we propose an estimation model for a resource plan. The model takes the performance characteristics of a physical machine as input and estimates virtual machine configurations. Using the estimated configurations, a user can obtain a virtual machine having similar performance characteristics to that of the physical machine. In this paper, we derive a model to estimate the CPU capability for creating a virtual machine on Xen. Our experiments show that our model can provide a resource plan for a virtual machine, which has the minimum performance difference compared with a designated physical machine.