The Flaw of PUE, A Single Number To Hide Behind

I am reading a lot on modeling and finding some good people to learn from.  One is Sam Savage a Stanford professor.

	Sam Savage Professor (Consulting) Management Science and Engineering

Research

* Fields of Specialization:
Embedding analytical techniques in spreadsheets, data bases and the OLAP environment, Risk Minimization in Petroleum Exploration, Stochastic Modeling in Accounting and the Law.

He has an article from Oct 8, 2000 in the SJ Mercury.

Published Sunday, October 8, 2000, in the San Jose Mercury News

The Flaw of Averages

If you count on the stock market's average return to support you in retirement, you could wind up penniless

BY SAM SAVAGE

``The only certainty is that nothing is certain.''

So said the Roman scholar Pliny the Elder. And some 2000 years later, it's a safe bet he would still be right. The Information Age, despite its promise, also delivers a dizzying array of technological, economic and political uncertainties. This often results in an error I call the Flaw of Averages, a fallacy as fundamental as the belief that the earth is flat.

The Flaw of Averages states that: Plans based on the assumption that average conditions will occur are usually wrong.

A humorous example involves the statistician who drowned while fording a river that was, on average, only three feet deep.

One of the points Professor Savage makes is

While many of today's managers still cling tenaciously to ``flat earth'' ideals, the innovators are abandoning averages and facing up to uncertainty. Those who dare discover a New World of managerial tools including simulation, decision trees, portfolio theory and real options.

And what happens when one of these innovators is confronted by someone cloaking themselves behind a single number? The story of the emperor's new clothes says it all.

I am constantly amazed how many people hold up PUE as a single number.

The guys at Google publish their PUE as not just a single number.

Quarterly energy-weighted average PUE:	1.20
Trailing twelve-month energy-weighted avg. PUE:	1.19
Individual facility minimum quarterly PUE:	1.15, Data Center B
Individual facility minimum TTM PUE*:	1.14, Data Center B
Individual facility maximum quarterly PUE:	1.30, Data Center H
Individual facility maximum TTM PUE*:	1.22, Data Center A

* Only facilities with at least twelve months of operation are eligible for Individual Facility TTM PUE reporting

What we need are more graphs showing the range.

Figure 2: Daily average PUE data for a new Google data center currently in bring-up

Comments

[Jack Pouchet]

Good that you pointed out the benefit of understanding a data center’s PUE on a more dynamic basis and not on a wide time-frame average basis. To be of value in gauging the on-going performance of a data center trending both the PUE along with the actual useful work performed against total energy consumed will provide valuable insights into data center productivity. PUE alone has nothing to do with productivity and suffers from the fallacy of assuming all IT loads including idle servers, networks, and storage systems have value.

We strongly encourage the data center community to work with The Green Grid to develop a set of proxies for useful work in the data center so we can have a more meaningful discussion centered on productivity.