Why don’t we see more energy saving activity than just talk loudly about it?

I’m sure that we have all walked round them: Data centres without blanking plates and hole stopping – even without cold-aisle/hot-aisle layout and quite rarely with aisle containment. It can’t be because there is any doubt about the energy saving available with RoI’s of less than 6 months, can it? Definitely not.

Then there are all the other things that can be done, even in a traditional facility loaded at 60 percent, to get the average PUE₀ from 2 to 1.5, e.g. raising the server inlet temperature to 27⁰C, raising the chilled water temperature so as to increase free-cooling operation, enabling eco-mode in the UPS (if they have it, which most have). Then you can reduce energy consumption (albeit often with a negative effect on PUE) by turning off comatose servers (not forgetting to blank them off with gaffer tape!) and virtualising the hell out of what’s left. Or, for the maximum impact, refresh any ICT hardware that is more than three years old and see the kW tumble…

So, if it’s more than possible to reduce energy consumption by a quarter just in terms of PUE (at low cost) and maybe 100 percent by a total ICT refresh (although that is a huge investment) why doesn’t ‘everybody’ do it? I have heard many presentations by the ‘green’ IT projects that are funded by the EU (with tax-payers money that is probably a better investment than funding inefficient farming and worrying about the straightness of bananas!) where the conclusion is that ICT users are somehow ‘crazy’ not trying to save such a lot of money.

But not one of these presentations (to date) have accepted that energy saving is NOT highest on the agenda of data centre operators – that honour is still reserved for reliability and availability of digital services, as it has always been. A chum in London runs a big data facility and put it to me rather succinctly as ‘I won’t get a bonus or promoted for saving on energy but I will get fired for losing the load’.

But what is the order of magnitude of business value (sales turnover) that relies on the data centre unit of energy (kWh)? Actually it is not that difficult to (very) broadly estimate an average for the UK based on a few data points and estimates:

• UK GDP is £2,000 million and Services represent 78 percent. If we assume that 80 percent of that business is dependent upon ICT we can reach an estimate for the business value that depends upon the digital infrastructure.
• Data centres consume three percent of the grid capacity, which averages out at 45GW, so we have an estimated energy that feeds data centres.
• Using an optimistic PUE of 1.50, if you do the arithmetic it turns out that 1kWh of ICT load in a UK data centre enables circa £100 of business and, critically, at an energy cost of £0.10/kWh, the power consumed represents less than 0.15 percent of the business value.

Therefore should we be surprised if energy cost is not of paramount focus on most user’s radar? Yes, saving maybe 25 percent on data centre power is attractive in isolation, but not by increasing risk of service failure and negatively impacting business of c700 times the cost of the power. A 25 percent reduction in data centre power would be a huge achievement but is only a 0.04 percent impact on the business cost.

For a reality check we can take one average cabinet of 5kW (a fully configured blade chassis) which, on this basis, enables £4.5m of business turnover by consuming £6.5k of electricity. However, clearly, some business models rely on data centres more than the ‘average’ and whose turnover is huge – for example banking and finance – but these only serve to reduce the impact of energy cost on the behaviour/attitude about taking risks.

Another example would be the UK Revenue (HMRC), a system totally reliant on ICT, the Internet and data centres. For that ‘business’ the money spent on data power is, literally, inconsequential with respect to the tax collected – I estimate less than 0.001 percent. With this driving operational behaviour and the need for reliability and availability being overriding it is not surprising that the facilities we see do not take risks with temperature, humidity, air-quality and power-quality.

Now, I am NOT saying that we should not do all we can to reduce energy, especially since demand for digital services continues to outstrip the technology curve, but we should be realistic about our expectations of rapid behavioural change in data centre users.