WP#72 - ICT Capacity and Utilization Metrics

Overview

The Green Grid's white paper number 72 is dated 23 January 2017 and is the product of an initiative that started in early 2015. It has a large author group that includes Rona Newmark, Kathrin Winkler, Jay Dietrich, Dale Sartor and Mark Seymour. Access requires signing in with a free account, and copyright rests with The Green Grid. This record rests on the abstract on the public landing page; the full text has not been read.

The paper defines two metrics: ICT capacity is the maximum volume of ICT service that can be provided for an investment, and ICT utilization is the share of that capacity actually used. The pair helps ask not only how much energy a facility uses but how much work that energy delivers. Where PUE measures infrastructure efficiency, these metrics focus on how well the ICT hardware is used.

Utilization is treated as a proxy for energy efficiency, because ICT equipment is more efficient at higher utilization. The abstract reports, citing SERT analysis, that doubling utilization can be achieved with roughly 4-5% more energy. This points to how costly idle hardware is in energy terms. The metrics deal with average utilization. This is why consolidating lightly used servers is an effective route to better energy efficiency.

As a limit, averages can hide peaks and troughs; the real load profile may differ from the mean. The record carries no calculation formulas or example values for the metrics. The source gives designers a conceptual framework for adding a capacity and utilization layer on top of PUE; the full text of the report must be consulted for calculation details.

Key points

1. Dated 23 January 2017; the initiative started in early 2015.
2. Defines ICT capacity and ICT utilization metrics.
3. Doubling utilization can be achieved with roughly 4-5% more energy (cited from SERT analysis).
4. The metrics deal with average utilization.
5. Because ICT is more efficient at higher utilization, utilization is a proxy for energy efficiency.
6. Averages can hide peaks and troughs.

Using it in design

• Add an optional IT-capacity panel: store compute, storage and network capacity and average utilisation per hall.
• Use this data to estimate useful-work efficiency beyond PUE.
• Note in reports that average utilization can hide peak loads.

Who should read it

Data center designers, operators and IT planners who want to measure ICT efficiency beyond PUE.

Publication details

Authors

Rona Newmark; Kathrin Winkler; John Calhoon; Charles Carroll; Dan Comperchio; Pierre Delforge; Jay Dietrich; Michael Dunbar; James Espy; Daniel Gaffney; Donald Goddard; Magnus Herrlin; Phil Isaak; J Carlos Pratt; Leonid Rabinovich; Dale Sartor; Mark Seymour; Roger Tipley; Todd Traver; Gary Verdun; Jay Vincent; Henry Wong

Publisher

The Green Grid

Year

2017

Document number

WP 72

Document language

English

Source type

Industry body

Access

Citation only

Licence

Copyright retained; citation only — the publisher’s terms apply to the original document

Citation