A review on evaluation metrics of thermal performance in data centers

Abstract

The rapid development of cloud computing and big data has enabled large-scale applications of data centers over recent years. Characteristics and demands of the operation of data center account directly for its high energy consumption level, making thermal management and air distribution performance being significant for reliability and energy saving for data centers. To evaluate thermal performance of the data center, various indicators have been proposed. This paper reviews the state-of-the-art of popular existent evaluation metrics. The main focus is on their characteristics and scope of application. Based on the applicable level, the data center is divided into room level, row level, rack level, and server level to analyze the application of metrics. Main advantages and limitations of major metrics are analyzed critically in order to suggest operators to select appropriate evaluation indexes for the design and optimization of data centers according to their specific building condition. Thereafter, a concept diagram of thermal performance of various levels is proposed for future work. Results show that evaluation metrics are valuable to real time feed-back on air management, allowing for scalability across rack, row and room levels. But still, more comprehensive and multi-scale evaluation metrics of the thermal performance of data centers are in need to combine reliability and energy savings in evaluation.