Why data centre energy consumption is so contentious
It can be said that data centre energy use is a bit of a Donald Rumsfeld conundrum. There are known knowns, known unknowns and unknown unknowns. Anyone claiming to be in possession of accurate figures for aggregate sector energy use is likely to have a limited understanding of the facts or, more worryingly, a political or commercial agenda.
However, lack of data has not prevented people from making confident statements. There have been some shockingly, perhaps wilfully, inaccurate estimates over the last few years. Here, Fryer looks at the problem and helps bring some realism to this space.Why is estimating data centre energy use so tricky?
Putting a figure on our aggregate energy use is complicated for a multitude of reasons. Firstly, the sector is a mixture of commercial operators, managed service providers and enterprise. While commercial providers are visible and relatively easy to account for, in-house operations are often under the radar so we have to find the facilities before we can add up their energy use.
Secondly, business models are complex and it can be tricky to segregate the data centre operation from the multiple levels of service wrap. Thirdly, the speed of change in the sector means that calculations, and especially assumptions, are very soon out-of-date.
Because of this uncertainty, we should expect to see vague and highly conditional figures with large error margins and all kinds of caveats, but that is not what is happening. Instead, we get simplistic, poorly or un-referenced claims for data centre energy use.
Most worryingly, the trend is to exaggerate energy consumption, sometimes to epic proportions. The most dramatic figures are quoted and re-quoted and have generated an elaborate mythology.
Why do we overestimate power use?
- Definitions: the way we define the sector, inevitably affects the result. Data centres consolidate IT functions into secure facilities with controlled environments and guaranteed power and connectivity. Server cupboards and other distributed IT should be excluded from calculations. If included this should be explicit.
- ‘Give me a number – any number’: the desire for a simple number pressurises people to plump for random figures, which are frequently taken at face value without checking context or acknowledging the assumptions made. Circular referencing also plays a part here.
- ‘Go on, prove me wrong’: it is hard for the industry to disprove claims categorically. Instead, arguments have to be focused on methodology and assumptions being wrong.
- Policy agendas: policymakers wishing to justify regulation exaggerate the energy consumption of the sector to create greater potential savings from interventions. This is unfortunately very common.
- Commercial agendas: we have seen a similar tendency among those providing energy efficiency solutions to overstate power use to maximise the potential savings.
- Failure to apply common sense: if your figure for data centre energy consumption represents 93% of the UK’s total generated supply then it’s unlikely to be correct.
- The Russian doll syndrome: data centre operations are multi-layered with energy resold to tenants and sub-tenants. Each member of the chain reports the same energy, multiplying the real figure. This is a particular problem in surveys.
- Confusing power provisioning with energy consumption: power provisioning is the instantaneous, maximum power supply available to operators. Multiplying it by 8760 to convert Watts to Watt-hours would give a massively bloated figure. Supply cannot be used as a proxy for energy use: it is not the same thing.
- Why let the truth get in the way of a good story?: “We’re all going to live in caves and eat bugs because data centre power use is out of control” is a much more exciting headline than “Data centres are using slightly more power but increasing productivity faster.”
A better way
The bad news is that governments and other stakeholders have an unfortunate tendency to understate the economic contribution of data centres while overstating their energy consumption.
The good news is that a body of real evidence is gradually building up and this means that we are chipping away at the known unknowns and unknown unknowns.
At Data Centre World in March, Fryer will outline some recent third-party studies that have shed more light on energy use. She also plans have a bit of fun with a few shockers – claims made by people who really should know better.
She will also explain how the data we get from the Climate Change Agreement is critical in informing this discussion because the 2.5TWh a year consumed by the UK’s commercial providers is one of the very few genuine known knowns.
In the meantime, those who make unqualified claims about data centre energy use will continue to do so. Our job is to question their figures, the motives behind them, to ask about sources and assumptions and then, where necessary ‘Name and Shame’.