The European Union is keen to measure the energy efficiency of the power systems inside data centres. Is it looking at the right things?
In 2013 the European Commission launched a pilot scheme likely to lead to a new set of regulations intended to reduce the environmental footprint of data centres. The pilots aim to develop an EU-wide standard measurement of the environmental performance of products throughout their life cycles, from shoes to solar panels.
The process will lead to what the Commission calls Product Environmental Footprint Category Rules (PEFCR), driven by the wider directive 'Building a Single Market for Green Products'. Although many of the pilots are concerned with foodstuffs and consumer goods, several are focusing on issues that affect data centres, particularly the environmental cost of uninterruptible power supplies (UPS), with others looking at servers within the wider context of IT equipment and the batteries that may go into the UPS subsystems.
Within the data centre, UPS products provide emergency power when the mains power fails. How a UPS differs from an auxiliary or emergency power system or standby generator is that it uses energy stored in batteries, supercapacitors or flywheels to provide near-instantaneous protection from input power interruptions. Although the on-battery runtime of most UPS products is relatively short, only a few minutes in most cases, it's enough to start a standby power source or properly shut down the protected equipment.
With the boom in cloud computing, software as a service and the rapidly growing Internet of Things industry, data centres are becoming an ever more important part of the IT landscape and the UPS is vital to ensuring that the data they store on behalf of customers is not lost to a sudden power outage. The economic importance of these centres mean they accounted for 3 per cent of the world's electricity consumption in 2013. And the share is growing.
Microsoft now has 19 data centres worldwide that can each fit two jumbo jets in them, Google currently has 12, and Apple recently announced that it will invest $1.9bn in its first European data centres, in Denmark and Ireland. So these global giants may well find that they'll have to carry out life-cycle assessments (LCAs) and abide by the PEFCR for the power supplies for their European operations, as well as the ISO standards that are already in place. The issue is not just about electricity consumption but the impact on resources of an industry that is currently upgrading systems at a rapid pace to take advantage of faster processors and memory. The processing blades that go into data centre are in the fastest growing part of the market for server hardware.
Life-cycle assessment is a framework tool used to measure the true environmental impact of products. So when you are using it to look at the environmental performance of data centres you have to consider every step of the life cycle of the end products used in it, as well as those that were used to build it, and dismantle it. LCAs take into account not just the energy consumed during product operation but the 'embodied' energy and resource consumption of manufacture, installation, decommissioning and recycling.
By using life-cycle analysis regulators believe that both vendors and users can have a more informed view of the environmental impact of the overall system. The result will be a more holistic view of the decisions that matter in the creation of green data centres and the IT equipment that goes into them. For example, a product with high operational efficiency but low lifetime reliability, or which is based on a less sustainable design, may not provide the basis for a greener product line. The update rate also has an effect. If a product has to be thrown away to make way for a faster processor instead of allowing partial upgrade, the overall life-cycle impact will be higher.
The International Organisation for Standardisation (ISO) has already standardised a framework of tools to manage LCA within the series ISO14040, so why do we need the new PEFCR?
The PEFCR difference
According to the European Commission, the PEFCR will "provide specific guidance for calculating and reporting products' life cycle environmental impacts". It believes that current guidelines "do not provide sufficient specificity to ensure that the same assumptions, measurements and calculations are made to support comparable environmental claims across products delivering the same function".
But the PEF pilots were launched and leading manufacturers are now providing information to help develop product category specific, lifecycle based rules for design and production, aiming to provide a more detailed and complete treatment of environmental impact than the calculations of operational efficiency that are done today.
Professor Ian Bitterlin of the University of Leeds and chair of the British Computer Society's data centre specialist group on server life cycle, says: "The proposals seem to be based on establishing the carbon footprint over the entire product lifecycle from manufacturing through to disposal. This is not an uncommon approach but it's just not that productive for servers because the end-result will prove that the biggest 'energy' impact is in the use phase and certainly swamps the embedded energy. The packing, delivery and recycling phases will be found to be inconsequential."
Bu the specific element of data centre performance that the PEF appears to care about most is power supply, with UPS having its own dedicated category, whereas servers and their internal power supplies are just included in the broader IT equipment category, which also covers magnetic disk units and switching equipment and has group members made up entirely of Japanese companies.
Within the datacentre, the UPS has come under particular scrutiny. The life-cycle situation for UPS installations is more complex than that of servers and because of the losses associated with them. Google says one of the largest losses in terms of power distribution is from the UPS. UPS manufacturers are working on optimising the energy efficiency of their products at a number of levels. UPS sub-group members include representatives from CEMEP, the European trade association of UPS manufacturers, Emerson Network Power, Schneider Electric, Eaton and Socomec. The aim for the UPS group in the pilot is to design universal UPS technology solutions that maximise availability, efficiency and capacity within the data centre environment.
Work started on the PEF pilot in February 2014. Christophe Garnier, who is environmental standards and alliances manager at Schneider Electric and part of the PEF pilot team, says: "With new technology developments it's looking likely that the energy efficiency of UPS can be improved significantly."
Energy efficiency is a key issue, but the pilot is looking at the whole life-cycle impact. Garnier says market requirements and legislative regulation are already pushing manufacturers to consider environmental concerns during all phases of the product."
The first stage of the UPS pilot, the screening study, has been completed. This saw the team create averages from data provided by UPS manufacturers on energy consumption and raw material use and assess its environmental impact. Preparations are now being made for the second stage, where each manufacturer will use the EU Commission methodology and the outcome of the screening study to make life-cycle assessments of a real UPS. Results are expected to be available late in 2015.
Greener server design
Energy efficiency is always a buzz phrase bandied around when discussing the environmental performance of anything but Bitterlin doesn't think it should be applied in the context of data centres and servers: "There is no such thing as 'efficiency' of' ICT hardware as it has no 'work' output to be compared with input energy. All that comes out of a server is 100 per cent heat and digital services. The ISO/IEC 30143 standard, which has developed KPIs for resource effective data centres, has addressed this descriptor problem by using the word 'effectiveness' in place of 'efficiency'."
Another way of looking at this is Carbon Usage Effectiveness (CUE), a metric that will appear for the first time in ISO/IEC 30134, which is currently in development and looks at data centre performance under the broader category of 'sustainability for and by information technology'.
A lot depends not on the data centre but the local generating infrastructure, says Bitterlin: "In the UK we have a relatively low proportion of renewable and low-carbon electricity in our utility, so the carbon footprint of a server is high. In France they have a high proportion of nuclear, which some people don't like but it's very low in carbon. In Scandinavian countries they have hydro, with Norway having almost 100 per cent renewable energy in their grid."
Moreover, energy costs can be minuscule in relation to the digital business handled by the server. This explains why end-users may not place energy effectiveness above service availability: "the cost of powering a server is minor compared to the losses of service disruption," Bitterlin says.
The issues created by the fast-growing energy demands of the ICT sector, combined with the fact that it also has a key role to play in making our buildings, transport, agriculture and public services more sustainable, is being looked at by another EC project. ICT for Sustainable Growth is focused on developing a more sustainable industry and has funding for research and development projects that look at how to improve the energy consumption and carbon footprint of ICT products and services.
Bitterlin believes one big area to address is utilisation: "The biggest problem in the broader ICT estate is lack of hardware utilisation. The average server is 10 per cent utilised and at that load consumes 40 per cent of its peak power."
Utilisation is what matters
Currently, the environmental performance of data centre infrastructure is measured by Power Usage Effectiveness (PUE) - the 'overhead' in energy that is required to run the ICT hardware. Bitterlin says: "We already know how to get this down to 1.1 to 1.3 compared to 2 to 3 of just five years ago and we also know that it is minimised by full load in the facility.
"The problem with PUE is that people confuse it with a data centre 'goodness' metric. It is not. A data centre can have a PUE of 1.2 but be wasting 90 per cent of the total energy because the utilisation is low. Improving server effectiveness is the only way to improve data centre effectiveness."
So, there are many issues to be addressed in making data centres greener and a study looking at the lifecycle environmental costs of UPS and servers is just one very small element of a much wider problem.