Data Centers (DC) are complex systems, which provide an environment to host Information and Technologies (IT) equipment. DCs are typically composed of a variety of components including servers, storage systems, networking infrastructures in addition to non-IT equipment such as power distribution and cooling systems. In order for a DC to function properly, all of its components need to be correctly configured and integrated. However, the variety of configurations and the interdependencies of DC sub-systems create challenges in understanding and optimizing DC complexities. In this paper, we present a system modeling approach, which supports the design of DCs and a quantitative assessment of their energy efficiency. This modeling approach aims at controlling the design complexity of DC infrastructure and ensuring the consistency of such designs. The cornerstone of our approach is a generic DC metamodel (DCMM), which captures the DCs heterogeneous structure, main characteristics, and diverse constraints. Furthermore, we propose a method to generate a Power Flow Model (PFM) from an input DC model. The PFM is then used to compute the DC Power Usage Effectiveness (PUE) metric. We evaluate the applicability of our approach through a use case of the Consortium Laval-UQAM-McGill and Eastern Quebec (CLUMEQ) data center.
Model-based Approach to Data Center Design and Power Usage Effectiveness Assessment
Attachment: Download Attachment
About the Author: Xu Liu
Brain-inspired computing Deep learning Machine learning Embedded system Image processing