Power-Thermal Modeling and Control of Energy-Efficient Servers and Datacenters
Recently, the energy-efficiency constraints have become the dominant limiting factor for datacenters due to their unprecedented increase of growing size and electrical power demands. In this chapter we explain the power and thermal modeling and control solutions which can play a key role to reduce the power consumption of datacenters considering time-varying workload characteristics while maintaining the performance requirements and the maximum temperature constraints. We first explain simple-yet-accurate power and temperature models for computing servers, and then, extend the model to cover computing servers and cooling infrastructure of datacenters. Second, we present the power and thermal management solutions for servers manipulating various control knobs such as voltage and frequency of servers, workload allocation, and even cooling capability, especially, flow rate of liquid cooled servers). Finally, we present the solution to minimize the server clusters of datacenters by proposing a solution which judiciously allocates virtual machines to servers considering their correlation, and then, the joint optimization solution which enables to minimize the total energy consumption of datacenters with hybrid cooling architecture (including the computing servers and the cooling infrastructure of datacenters).