The study and analysis of energy efficiency in Data Centers (DCs), through a set of globally accepted metrics, is an ongoing challenge. In particular, the area of productivity metrics is not completely explored, and there is no existing proposed metrics, which provides a direct measurement of the useful work in a DC. This paper proposes a methodology that addresses the problem of measurement, calculating, and evaluating the energy productivity assessment in Data Center (DC), which encompasses both the portion of energy employed for computing and energy wasted during computational work. It involves the estimation of productive energy consumption by a DC cluster based on the following: statistical data collection and interpretation, software for energy data analysis, and mathematical formulation. This current work is based on available data extracted through experiments conducted on the cluster “CRESCO4” from ENEA Data Center facilities. The dataset covers the power and job schedule characteristics running on the cluster for one year. This paper shows how to advance beyond state of the art for productivity metrics (e.g. useful work). It will also help enhance server performance and power management since the appropriate statistical data analysis provides a profile on server energy consumption behavior. Additionally, we make recommendations on how the productivity assessment could driver a new power efficiency management strategy, which is specifically targeted at DC manager and/or operators, and end-users of the facilities.
DC Energy Data Measurement and Analysis for Productivity and Waste Energy Assessment
Autori
Anastasiia Grishina, Marta Chinnici, Davide De Chiara, Guido Guarnieri, Ah-Lian Kor, Eric Rondeau, Jean-Philippe Georges Parole chiave (Tematica)
big data Cluster Computer Application Data Analysis data center energy consumption energy efficiency metrics Policies Workload Management Data pubblicazione
27/12/2018 Fonte
2018 IEEE International Conference on Computational Science and Engineering (CSE)
Electronic ISBN: 978-1-5386-7649-3 Print on Demand(PoD) ISBN: 978-1-5386-7650-9