Capacity planning for virtualized servers

Today's data centres offer many different IT services mostly hosted on dedicated physical servers. Virtualization provides a technical means for server consolidation leading to increased server utilization. The term refers to the abstraction of computing resources across many aspects of computing and has been used to describe different techniques. Virtualization engines allow hosting multiple virtual servers (including operating system plus applications) on a single physical server, and sometimes to migrate and dynamically allocate these virtual servers to physical servers on demand. This allows for much flexibility in capacity management. This article presents a number of capacity planning problems in the presence of virtualized IT infrastructures and decision models to allocate these virtual servers optimally. Based on a data set of CPU traces from a data centre provider, we present the results of experiments using different problem formulations.