WP#64 - Data Center Segmentation: Workload Archetypes

Overview

The Green Grid WP 64 is dated 20 May 2015 and was written by Henry M. Wong (Intel). It defines workload archetypes that group typical data centre applications by the useful work they deliver.

The paper's starting point is that energy efficiency is defined as useful work per energy consumed. Since the useful work of one facility or application differs in nature from another's, using a common yardstick first requires splitting workloads into meaningful groups. Archetypes provide that grouping.

Archetypes describe typical workload and application groupings and enable standard comparison metrics that reflect different productivity targets, as well as baseline comparisons between facilities. The approach is an attempt to look beyond infrastructure-only indicators such as PUE toward the efficiency of the IT side.

As a limitation, the page does not include the full list of archetypes or any metric formulas.

Key points

1. Dated 20 May 2015; author Henry M. Wong (Intel).
2. Energy efficiency is defined as useful work per energy consumed.
3. Archetypes describe typical workload and application groupings.
4. They enable standard comparison metrics.
5. They enable baseline comparisons between facilities.
6. They lay the ground for indicators reflecting different productivity targets.

Using it in design

• At project setup, let users tag a facility with a workload archetype (for example HPC/AI, enterprise, cloud, storage).
• Select load profiles and benchmark ranges according to that archetype.
• Avoid comparing facilities of different archetypes directly in reports.

Who should read it

Engineers and strategists who want to assess data centre efficiency in terms of useful work, not only infrastructure.

Publication details

Authors

Henry M. Wong

Publisher

The Green Grid

Year

2015

Document number

WP 64

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