This past Thursday I received an email from [email protected], along with a copy of “Recommendations for Measuring and Reporting Overall Data Center Efficiency: Version 1”, as did, I suspect a million or so of my closest colleagues in the data center and related industries. The note referenced a January meeting of 7x24 Exchange, ASHRAE, The Green Grid, Silicon Valley Leadership Group, U.S. Department of Energy’s Save Energy Now and Federal Energy Management programs, U. S. Environmental Protection Agency’s ENERGY STAR Program, U.S. Green Building Council, Uptime Institute among others that produced the referenced document. Within minutes, I received the first of a steady stream of broadsides and inquiries and “FYI’s”. This first response is pretty much a sign of the times and people’s general confidence in institutions of any kind:
"Would somebody help me out here and explain to this amateur if I understand correctly. These eminent industry leaders got together for 2-3 days and came up with:
- When calculating PUE, IT energy consumption should, at a minimum, be measured at the output of the uninterruptible power supply (UPS). However, the industry should progressively improve measurement capabilities over time so that measurement of IT energy consumption directly at the IT load (i.e. servers) becomes the common practice.
- For a dedicated data center, the total energy in the PUE equation will include all energy sources at the point of utility handoff to the data center owner operator. For a data center in a mixed-use building, the total energy will be all energy required to operate the data center, similar to a dedicated data center, and should include IT energy, cooling, lighting, and support infrastructure for the data center operations.
That’s the earth shattering news? That the IT energy consumption should be measured at the output of the UPS?
That’s why they all had to get together? Probably took one day to come up with this and two days to expand to the final 12 page document."
Granted, it may be a little silly, but at the same time sometimes the simplest little thing can get complicated really fast. For example, if you are aspiring to operate at a tier 3 or tier 4 level, there are going to be most topologies where it really makes sense to have your air handler fans on UPS. That increases your divisor in the PUE equation, resulting in tier 3 and tier 4 data centers not receiving the appropriate penalty for the inherent excesses required to support the higher availability. And, lo and behold, the EPA was surprised by the results of their benchmarking studies that there was not a terribly significant efficiency difference between the different tier levels. Oh gosh!
Now, for the original purpose of PUE – i.e., benchmarking yourself for continuous improvement, it really doesn’t matter in the least. However, as the PUE metric is now the basis for determining who can achieve Energy Star rating status, it starts to have broader implications. For example, a data center that is operating wastefully at a higher tier level than is really mandated by the business purpose of the data center, will not be appropriately penalized for that wastefulness if they can stick it in the divisor rather than the dividend when they are making their calculations. I don’t know anything about the meeting referenced here, but I sat in on most of the EPA conference calls and this was a subject of ongoing debate and we all basically decided to leave the metric at the UPS output to avoid the additional measuring and monitoring complexity that, with today’s technology, would become so unwieldy it would in itself become an inhibitor to businesses’ participation.
So, while on the surface it seems like a pretty basic output from all the time and energy dedicated to it, it really represents a pretty significant compromise for practicality over accuracy. Furthermore, as with the decision to exclude a geographic handicap in the PUE calculations, these compromises probably will not cloud the program in the first year or two because there is so much room for improvement throughout the industry that it is pretty safe to say that the Energy Star ratings earned early on will more or less accurately represent the targeted 75th percentile of the industry.
For additional information visit the Energy Star Data Center Efficiency Website. Ian Seaton, Technical Applications Development Manager