Archived Content

The following content is from an older version of this website, and may not display correctly.

PUE is just one of the range of metrics that can positively impact data center effectiveness but users are in danger of misusing PUE or over relying on targeting simply low PUE numbers.

A number of metrics have been  developed by different contributors and they include: Data Center Compute Efficiency, Performance per Watt, Data Center performance per energy (DPPE) and Carbon Usage Effectiveness.

Derek Allen, Group Operations Director at Global Switch quoted from a list of at least 15 measurements that are being offered as metrics.

Global Switch use PUE to assess its data center effectiveness. “We believe there is a case for creating an industry PUE reporting standard calibrated for real use, because currently different organisations use different measurements to obtain PUE numbers and a consistency of approach is required,” said Allen.

In October this year Global Switch conducted a survey among end users and its findings revealed PUE is used and viewed inconsistently.

Global Switch categorised some examples of measurement methods being applied

§  PUE Cat 0: IT load measured at UPS output(s). Total dc power measured at the utility meters.
§  PUE Cat 1: IT load measured at UPS output(s). Total dc power measured at the utility meters. 12-month cumulative readings
§  PUE Cat 2: IT load measured at electrical distribution panels supporting IT loads.
§  PUE Cat 3: IT Load measured at the point of connection of IT devices to electrical system. Total dc power measured at the utility meters. 12 month cumulative reading
§  Use an alternative local method for measuring PUE
§  Other

21% of those surveyed said they used category 3 for measuring their PUE compared with 17% for Cat. 2 and 14% for Cat 1.  41% used other undisclosed measures.

Allen said “there is a danger of ‘hyped up messaging’ and the whole PUE discussion becoming populated with marketing driven sound-bites rather than factual repeatable data”.

Allen presented an illustrative hypothetical example of when good efforts can deliver negative results in PUE ratings.  An IT virtualisation project reduced the power drawn from IT equipment by 50% from 50kW to 25kW in an environment that originally drew 100kW (IT and facility power). Even if the input power dropped to 75kW, against a draw of 25kW by this positive action, this increased the PUE number from 2.0 to 3.0.

“PUE should be one of a number of metrics when reviewing data center performance or making comparative decisions and the industry should make efforts to calibrate PUE for real use, or we risk misleading end users and ourselves” said Allen.