Evaluating Dynamic Resource Allocation Strategies in Virtualized Data Centers

Virtualization technology allows a dynamic allocation of VMs to servers. It reduces server demand and increases energy efficiency of data centers. Dynamic control strategies migrate VMs between servers in dependence to their actual workload. A concept that promises further improvements in VM allocation efficiency. In this paper we evaluate the applicability of DSAP in a deterministic environment. DSAP is a linear program, calculating VM allocations and live-migrations on workload patterns known a priori. Efficiency is evaluated by simulations as well as an experimental test bed infrastructure. Results are compared against alternative control approaches that we studied in preliminary works. Our findings are, dynamic allocation can reduce server demand at a reasonable service quality. Countermeasures are required to keep the number of live-migrations under control.