Is PUE killing your competitive edge?

Data is big business, currently making up 8.3 per cent of the UK economy. Energy is big business, worth £28 billion in the UK alone. As energy prices continue to rise and demands grow exponentially on IT, it’s vital that data centre managers use both energy and data to their advantage to ensure their facility is the most competitive on the marketplace. Unfortunately, though many think that this is exactly what they’re doing, they’re wrong. They’re being misled by metrics. By Mike Jansma, Co-founder and Chief Marketing Officer, Enlogic.

10 years ago Posted in

THOSE WHO STRIVE to make their data centre as efficient as possible will not only combat the environmental impact of an increasingly IT reliant world, but will also enable them to pass significant energy savings onto their customers, turning them into a more attractive prospect than their less energy efficient rivals.

In their quest to become as energy efficient as possible, data centre managers are desperate for measurements that can effectively differentiate them from competitors. The metric used most often by data centre managers keen to brag about their energy efficiency is Power Usage Effectiveness (PUE). Invented in 2007 by the Green Grid, PUE is the ratio of total amount of power used by a computer data centre facility to the power delivered to computing equipment, and can be calculated using this equation:

While a useful measurement tool, a PUE calculation should never be quoted in isolation as an accurate illustration of a data centre’s overall energy usage. Differences in how and when power is recorded can cause large fluctuations in the accuracy of the PUE measurement. Further, the PUE measurement fails to recognise data centre energy efficiency improvements that may reduce total energy use, but actually raise the facility PUE. Therefore, PUE is a useful tool for facilities managers, but shouldn’t be the only way in which performance is measured in those data centres that want to remain competitive.

Glaring oversights
Data centre managers are being blindly driven by the belief that PUE is the magic metric. It’s not. It’s useful, but only when used in combination with other factors to measure performance.

To illustrate how glaring oversights in using PUE are made, take a look at a data centre that has 1,000 servers, but is only actively using 100 of those. So, to improve energy efficiency a data centre manager would be wise to turn off those additional unused 900 servers. In doing so a data centre manager unquestionably makes their data centre more energy efficient, but the PUE will go up, and by quite a lot, because the power being used is now comparative to 100 servers, rather than 1,000.

Studies by McKinsey & Co and Gartner Research revealed recently that a mere six to twelve per cent of the electricity consumed by most data centres is used to perform useful IT computations. It would seem that many data centres are responding to demands for service reliability and security by diverting electricity to ensure availability of idle servers in fear of unforeseen downtime. Each of them will be able to show a very good PUE measurement, but the environmental wastage of using electricity to power unnecessary equipment should be unacceptable in the modern competitive market.

The potential energy and financial savings that data centres could make are entirely overlooked by a measurement like PUE, and the industry’s recent fervent fixation on the metric has caused many to miss some much more transformative methods of measurement. However despite this, there are some forward-thinking data centre managers who realise that intelligent solutions can step in and play a critical role in boosting energy efficiency in the data centre. Historically, ‘dumb’ solutions have only provided straightforward power connectivity and routing for all the IT components. However, newly developed intelligent solutions not only understand which servers are idle, but also offer data centre managers the option of switching individual or group outlets on or off in response to real time needs. This removes the need for permanently switched-on stand-by, massively reducing energy consumption and creating a far more cost-effective alternative to a complete data centre retrofit.

Performance is key
More progressive companies are now looking not to single metrics, but rather to their overall performance in comparison to the energy they consume. Their goals are now to obtain the maximum level of performance for the minimum level of energy consumption.

One such company is internet shopping goliath Amazon. Rather than trying to reduce the PUE of its data centres, it’s instead measuring how many transactions it can process per watt hour of energy it uses. By improving this overall measurement, Amazon is taking a more holistic and effective approach to increasing its energy efficiency. Companies such as eBay and Google are also adopting similar approaches by measuring how many auctions and searches they can process per watt hour of energy used.

PUE is without a doubt useful, but only when used in context and shouldn’t be considered as the default tool that will paint an accurate picture of your data centre’s energy usage. It’s impossible to compare a data centre’s performance against its rivals by quoting its PUE as these figures are incomparable.

Those who only quote their PUE will be missing a trick in a competitive situation as they’ll clearly be under - or over-selling their data centre’s performance. Data centre operators who don’t appreciate the importance of accurately managing and measuring their data centre energy usage will fall behind those who do. The reality is that some may find themselves moved aside by data centres operated by forward-thinking managers who want to be part of the next generation data centre landscape.