Modelling Energy Efficient Server Management Policies in PEPA

Abstract

Power inefficiency has become a major concern for large scale computing providers. In this paper, we model turning servers on and off to keep a balance between capacity and energy saving. Several heuristic-based switching policies are introduced with a view to balance the cost between power saving and performance. Models are specified using a Markovian process algrebra, which allows explicit representation of system behaviour and facilitates numerical analysis using the supporting tools.