Energy consumption of today’s data centres is an important issue in terms of both costs for the data centre owner, as well as the impact that such consumption has on the environment. One suggested means of addressing this issue is to turn off servers in the data centre when they are not required, however this then leads to evaluating a delicate tradeoff between performance (response time of an arriving request) and energy consumption. The focus of this talk will be on presenting two models that capture the tradeoffs in such problems. The first is a multi server stochastic (queueing) model, which allows one to capture some of the inherent randomness (in interarrival and processing times). We discuss optimal policies both for systems with a finite number of servers and describe asymptotics as the number of servers grows large. The second model is deterministic and strives to capture a more detailed thermal model of a data centre. We will be mainly concerned with presenting the model itself and initial thoughts on optimization (this topic is work in progress).