Date Added: Apr 2010
The cloud computing paradigm proposes a flexible payment model, where a cloud user pays only for the amount of resources used. Thus, the cloud provider must only allocate as many resources to a user, as are to meet client performance requirements. In this paper, the authors present a control-theoretic approach to tune the CPU resource allocated to a Virtual Machine (VM) such that application performance metrics are optimized while using minimal CPU resource. Moreover, the approach expects no inputs of performance targets to steer the system towards optimal values of CPU share allocations and system performance. They prototype and validate the methodology on a small virtualized testbed using the Xen virtualization environment.