Resource
Metrics of IT Equipment — Computing and Energy Performance
This report, by Bruce Nordman of Lawrence Berkeley National Lab’s High Tech Buildings Project, summarizes work on the development of a simple, standard method of characterizing the degree to which a single server reduces its energy consumption when operating at low levels of computation compared to what it consumes at peak computing capacity (the “part-load” condition). The goal is to bring more attention to the issue, leading to future servers which save energy by having lower power use at part load.
This 22-page technical report is likely to be of use to data center managers and others in the server industry.