Intelligent Power Management Over Large Clusters

There is a growing tension within large organisations such as universities between the desire to perform vast amounts of computational processing and the desire to reduce power consumption by switching off computers. This situation will only worsen as computational problems get larger and the desire to save energy escalates. Through careful management of computing resources it is possible to maximise effective computer usage whilst minimising power consumption though this can be costly in terms of human effort. We present our work with the Agility Cloud Computing Platform to provide intelligent control over a University-wide Condor system, which works to reduce power consumption without adversely affecting the Condor users. This system also provides auditing of the power usage, which can be used to determine the power efficiency of the Condor system.