Efficient Management of Cooling Systems in Green Datacenters

Ibrahim Safieddine, Noël De Palma
Next-generation green datacenters were designed for optimized Power Usage Effectiveness (PUE), which is the ratio of the total power consumption of the datacenter over the computing consumption. Continuous improvement of these datacenters target the reduction of PUE in accordance with the servers' load, to approach the minimal target, i.e., PUE = 1. Datacenters must ensure in the same time a very high level of availability of resources and good management of failures, thanks to redundant
more » ... nts, while optimizing power consumption and cooling costs and reducing the environmental footprint. A datacenter consists of computing, power distribution and cooling parts. The cooling part represents the main cost that significantly increases the power bill and consequently the PUE. A datacenter can be cooled using heterogeneous cooling systems for redundancy in the case of failure. These systems have a variable consumption depending on the load and on external parameters, e.g., weather and external temperature. This paper presents an efficient cooling manager, which aims to minimize the PUE while satisfying the Service Level Agreement (SLA), by reducing the power consumption of the datacenter and opting for the most efficient cooling system according to climate conditions and by limiting temperature variations and cooling mode transitions. Our system uses an overview of all datacenter parts to provide an optimal decision that complies with regulations when using natural resources, e.g., groundwater.