On Energy- and Cooling-Aware Data Centre Workload Management

The power consumption of a data centre (DC) can be attributed to the power consumed for running the servers and to the computer room air conditioner (CRAC) power for cooling them. The challenge is to distribute the load among servers, controlling the number of active servers and optimally balancing IT and cooling power requirement. This goal demands integration of thermal, power and workload models that minimises a non-linear energy utilisation function. In this paper first we encode this problem using a non-linear objective function and use local search for solving it. We carry out simulation experiments using data provided by the Bluesim tool. The results encourage the effectiveness of our approach, showing that system-wide energy utilisation can be reduced using a holistic approach.