I see that Emerson US have proposed that a new KPI should be introduced to ‘capture the energy used in water consumption’. They appear to want to combine PUE with WUE. With PUE abuse still rampant by marketers and with its most vociferous ‘technical’ critics having never read the detail it is hardly likely that a new KPI will find an audience other than in a comedy routine. It is even more puzzling that the proposer knows full well that The Green Grid introduced both PUE and WUE and that all of the energy used in water consumption within a data centre (including the embedded energy of city-water) is already captured within PUE – if only people would take it seriously, which, of course, many fail to.
It was four years ago, when at Ark, that I wrote a white paper detailing the link between PUE and WUE – basically as WUE goes up your PUE goes down, although then I called it WUF with ‘F’ for factor [Traces of the paper exist - there’s a page mentioning it in our archives - Editor].
Use water yourself, and cut utilities’ water consumption
That paper went on to show that, in a UK type electrical utility, as you consumed more water locally the overall water consumption (power-station to data centre) fell quite dramatically as electrical generation in thermal stations uses vast quantities of water – albeit not potable. So the conclusion was using water in an evaporative system uses 1000 cubic meters per MW of load per year - equivalent to about 20 domestic houses. The impact on the pPUE reduced it from a target 1.4 to 1.08 and if you didn’t use potable water (e.g. rainwater harvesting) you avoided the embedded energy proxy, equivalent to about 0.008 in the pPUE. All of the energy for pumping, spraying, treating, flushing etc is included ‘naturally’ in the PUE. So, all in all, there is no need for a combination metric – just a need for formalising the embryonic WUE.
Does anyone care?
And there we have an interesting problem. I sit, amongst others, for the UK BSI on the committee drafting ISO/IEC SC39 KPIs for Resource Effective Data Centres. That already includes PUE (with the rights given up by TGG) and ‘intends’ to include WUE – except that TGG doesn’t want to let go of it yet. So, not to be deterred, as a member of TGG I volunteered to chair a Water Interest Group (WIG) to develop WUE a bit further (in my mind to prepare for a ‘Standard’) and kicked off by splitting WUE into 10 separate topics and writing a draft statement on each. I (and my co-chair) published these 10 documents on TGG web platform and expected the 40+ members of the WIG to comment away. The intension was then to forge out a WUE definition with meat on the bones. I was very disappointed that there was only one response other than from my co-chair. In particular the board member organisation that didn’t want WUE to be let go yet didn’t offer any comment at all. So, was I wrong to conclude that TGG aren’t really interested in water?
But is anyone interested in water consumption? A data centre doesn’t have to consume water at all – look at 99.9 percent of European facilities that use air-cooled chillers but clearly the topic is growing as we see that adiabatic and evaporative systems can make a very positive impact on the PUE (actually the ‘power bill’ is a more accurate descriptor).
I think at the heart of the problem (apart from greenwash politics and ‘sustainability’ nonsense of some large companies) is encapsulated in one of those ten sub-topics that we wrote; comparative consumption. Oddly the only comment that we got from the WIG of TGG was that they wanted to remove the concept of relative consumption. The objector didn’t really explain why they didn’t like it – especially as it is so generous towards data centres.
Data centers aren’t the problem
Look at Coca-Cola; they have a great reputation for minimising water consumption in soft-drink production but still use 2.5L of potable water to produce 1L of fizzy pop. If Coca-Cola reduced consumption by less than 1 percent the saving would be enough for all the data centres in the world to be indirectly adiabatically cooled (excepting those in humid regions, of course). Or consider the waterless urinals that McDonalds have installed in many restaurants; each one saves about the same water that 1MW of adiabatically cooled data centre would consume.
As a society we consume vast quantities of water and, yes, we should try to consume less, but the consumption in data centres is minuscule in comparison to the overall and has the added advantage of reducing electrical demand. Do we need a combined PUE and WUE? No, we need both to be separately used correctly.
What about WUE as a part of ISO/IEC 30134 series? TGG have the title to WUE but I think that the time has come to write the draft under the TLA of ‘WCF’ rather than WUF. That should include the impact in the local electrical utility. Why, Water Consumption Factor, WCF? Because Woof sounds daft…