Aiming at improving the energy efficiency of air cooled servers in data centers, we devise a novel control oriented, nonlinear, thermal model of the servers that accounts explicitly for both direct and recirculating convective air flows. Instrumental to the optimal co-design of both geometries and cooling policies, we propose an identification methodology based on Computational Fluid Dynamics (CFD) for a generic thermal network of m fans and n electronic components. The performance of the proposed modelling framework is validated against CFD measurements with promising results. We formalize the minimum cooling cost control problem as a polynomially constrained Receding Horizon Control (RHC) and show, in-silico, that the resulting policy is able to efficiently modulate the cooling resources in spite of the unknown future computational and electrical power loads.
Energy savings in data centers: A framework for modelling and control of servers’ cooling
Varagnolo D.
2017
Abstract
Aiming at improving the energy efficiency of air cooled servers in data centers, we devise a novel control oriented, nonlinear, thermal model of the servers that accounts explicitly for both direct and recirculating convective air flows. Instrumental to the optimal co-design of both geometries and cooling policies, we propose an identification methodology based on Computational Fluid Dynamics (CFD) for a generic thermal network of m fans and n electronic components. The performance of the proposed modelling framework is validated against CFD measurements with promising results. We formalize the minimum cooling cost control problem as a polynomially constrained Receding Horizon Control (RHC) and show, in-silico, that the resulting policy is able to efficiently modulate the cooling resources in spite of the unknown future computational and electrical power loads.
Pubblicazioni consigliate
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
