The increasing demand for computation and the commensurate rise in the power density of data centers have worsened the cost associated with constructing and operating a data center. Exacerbating such costs, data centers are often over-provisioned to avoid costly outages associated with the potential overloading of electrical circuitry. However, such over provisioning is often unnecessary since a data center rarely operates at its maximum capacity. Moreover, the inability of the servers to exhibit energy proportionality, i.e., provide the same energy efficiency under all levels of utilization, diminishes the overall energy efficiency of the data center. It is imperative that we realize effective strategies to improve the energy efficiency of data centers.
 
Power and resource provisioning techniques can be used to improve energy proportionality. However, introducing constraints on power and resource available can cause unacceptable violation of service-level agreements (i.e., throughput and response time constraints).

In this talk, I will describe my investigation into whether it is possible to achieve energy proportionality while meeting strict performance constraints. First, I will present an analysis of the average and instantaneous power consumption of different components in the system, their energy proportionality and opportunities for saving power. Next, I will describe a runtime system, based on a load prediction model and an optimization framework, for power and resource provisioning to improve energy proportionality.