The following posting originally appeared on Michael Manos' Loose Bolts blog.
Matt Stansbury over at Data Center Facilities Pro posted an interesting post regarding a panel containing Uptime’s Ken Brill. The note warns folks on the use of PUE as a benchmarking standard between data centers.
I can’t say I really disagree with what he says. In my mind, self measurement is always an intensely personal thing. To me, PUE is a great self-measurement tool to drive towards power efficiency in your data center. Do you include lighting? Do you include your mechanical systems? To me those questions are not all that dissimilar to the statement, “I am losing weight”. Are you weighing yourself nude? in the your underwear? with your shoes on?
I do think the overall PUE metric could go a little farther to fully define what *MUST* be in the calculation, especially if you are going to use it comparatively. But those who want to use this metric in some kind of competitive game are completely missing the point. This is ultimately about using the power resources you have to its highest possible efficiency. As I have stated over and over, and as recently as the recent Data Center Dynamics conference in Seattle. Every Data Center is different. If I tried to compare the efficiency of one of our latest generation facilities in San Antonio or Dublin to a facility built 10 years ago, assuming we made sure that we were comparing apples to apples with like systems included, of course the latest generation facilities would be better off. A loss of 5 pounds on an Olympic runner with 4% body fat compared to a loss of 5 pounds on professional sumo wrestler have dramatically different effects (or non effects).
Everyone knows I am a proponent of PUE/DCiE. So when you read this understand where my baggage is carried. To me the use of either, or, or both of these is a matter of audience. Engineers love efficiency. Business Managers understand overhead. Regardless the measurement is consistent and more importantly the measurement is happening with some regularity. This is more important than anything.
If we are going to attempt to use PUE for full scale facility comparison a couple of things have to happen. At Microsoft we measure PUE aggressively and often. This speaks to the time element that Ken mentions in his talk in the post. It would be great for the Green Grid or Uptime or anyone to produce the “Imperial Standard”. One could even think that these groups could earn some extra revenue by certifying facilities to the “Imperial PUE standard”. This would include minimum measurement cylces (once a day, twice a day, average for a year, peak for a year, etc). Heavens knows it would be a far more useful metric for measuring data centers than the current LEEDS certifications. But thats another post for another time. Seriously, the time element is hugely important. Measuring your data center once at midnight in January while experiencing the coldest winter on record might make great marketing, but it doesnt mean much.
As an industry we have to face the fact that there are morons amongst us. This, of course is made worse if people are trying to advertise PUE as a competitive advantage due mostly to the fact that this means that they have engaged marketing people to “enhance” the message. Ken’s mention of someone announcing that their PUE of .8 should instantly flag that person as an idiot and you should hand them an Engvallian sign. But even barring these easy to identify examples we must remember that any measurement can be gamed. In fact, I would go so far as to say that gaming measurements is the national pastime of all businesses.
Ultimately I just chalk this up to another element of “Green-washing” that our industry is floating in.
Ken also talks about the use of the word “Power” being incorrect and that because it is a point in time measurement versus an over time measurement and that we should be focused on “Energy”. According to Ken this could ultimately doom the measurement on the whole. I think this is missing the point entirely on two fronts. First whether you call it power or energy, the naming semantics dont really matter. They matter to english professors and people writing white papers, but it terms of actually doing something, it has no effect. The simple act of measuring is the most critical concept here. Measure something, get better. Whether you like PUE, DCIE or whether you want to adopt “Energy” and call it EUE and embrace a picture of a sheep with power monitoring apparatus attached to its back, the name doesnt really matter. )Though I must admit, a snappy mascot might actually drive more people to measure. Just do something!
My informal polling at speaking engagements continues on the state of the industry and I am sad to say, the amount of people actively measuring power consumption remains less than 10% (let alone measuring for efficiency!), and if anything the number seems to be declining.
In my mind, as an end-user, the thrash that we see coming from the standards bodies and think tank organizations like Uptime, Green Grid, and others should really stop bickering over whose method of calculation is better or has the best name. We have enough challenge getting the industry to adopt ANY KIND of measurement. To confuse matters more and argue the finer points of absurdity is only going to further magnify this thrash and ensure we continue to confuse most data center operators into more non-action . As an industry we are heading down a path with our gun squarely aimed at out foot. If we are not careful, the resultant wound is going to end up in amputation.