WIREDGOV

A study was yesterday published on energy consumption of cloud services and suggested areas for policy intervention.

The European Commission has just launched a study commissioned from the Austrian Energy Agency on the energy implications of cloud services in Europe, both now and looking ahead.  By cloud, think data infrastructure.  The study has been a while in gestation, runs to some 290 pages and will take some time to digest.  On first impression, however, it appears to be rather a curate’s egg. The first things to strike me before a more detailed analysis are as follows:

• The authors seem have been selective in their references - ignoring important sources and studies like that conducted by the International Energy Agency – eg https://www.iea.org/reports/data-centres-and-data-transmission-networks but included unsubstantiated projections and scenarios.  That’s a common failing so to be expected. Even so, the reference to “exponentially (sic) growth in energy consumption due to the expansion of cloud services at a European level” is demonstrably untrue, whichever study you look at.
• More worrying, perhaps, is that this study promulgates the myth that a cupboard containing three servers under a desk on office premises and a wholesale colocation facility are both data centres and that they can be regulated in the same way, with the same instruments, and that the same improvement measures are relevant in both scenarios.  Anecdotal evidence, and more recently hard facts from the Eureca project results tell us that this is a fallacy.  (Just to remind you, the Eureca project studied 350 public sector data centres assessed across Europe. The average PUE was about 4, the average utilisation was about 20%, 40% of servers were over 5 years old and these consumed 66% of the energy but only delivered 7% of compute.  The initial objective was to provide a toolkit to help local authorities improve efficiency but the obvious conclusion was that many of these facilities would be better outsourced, in the interests of both energy and cost savings.)  We estimate, very approximately, that 2TWh of power are wasted in the UK every year by efficiency shortcomings in on premises server rooms, cupboards and small data centres.  That is two thirds of the consumption of the UK’s whole commercial data centre sector! The problem is that these are completely untouched by any of the multiple regulatory interventions that apply to commercial operators.  Even more staggering, pre-contracted, energy efficient alternatives exist – for instance in Crown Hosting, for those server rooms within public sector bodies, but are not adopted, despite a compelling business and environmental case. Tackling that should be high on the priority list.
• The review of practices looked pretty sensible, but then these presumably came from the industry.  More efficient cooling and use of renewable energy are no-brainers and the sector is demonstrating ambition and good practice in both areas.  Heat re-use is higher on the list than it should be and may be ASHRAE should have got a mention, but broadly little to disagree with here.
• Recommendations – these are a mixed bag and include a number of sensible proposals – such as greater transparency, which is something we have proposed ourselves (see Lost in Migration: Attributing Carbon to Cloud).  The sector also has exceptionally well developed certification schemes and standards for such a young industry – the key here is not to go reinventing wheels.  We all know that eco-labelling a data centre is going to be a total can of worms, however, there are schemes being developed that will provide signals to consumers regarding energy and environmental performance, so this is a space well worth watching carefully.

So the Commission has to decide which problem it wants to solve, and while I am, as ever, frustrated that studies of this kind continue to promulgate rather than dispel myths, that does not mean that there is no room for policy action.  I’m sure I agree with the Commission that the focus should be on the energy related issues that are not going to solve themselves, but we may choose to differ on how we define those!   I will now have a more detailed read of the study so that we can collectively consider any implications.  Expect and update in the next few weeks, but in the meantime, you can find the announcement and the study itself here:

https://ec.europa.eu/digital-single-market/en/news/energy-efficient-cloud-computing-technologies-and-policies-eco-friendly-cloud-marketEuropean