Datacentre PUE: Who's keeping score?

Creative calculations

A PUE rating is calculated by dividing the total power consumption of a datacentre by the amount of energy used by the IT equipment inside it, and the final result can be relatively easy to manipulate.

Ways of changing the results include failing to add the power consumed by the datacentre’s lighting systems or office functions in calculations, or by basing them on the designed efficiency figures of the IT equipment inside the facility, rather than its real-world performance.

Another way to obtain a favourable PUE rating is to ignore the Green Grid Association’s advice to use annualised energy consumption figures in calculations.

Instead, operators use figures from periods when the IT inside their facilities is running at full capacity, as this is likely to return a lower PUE than if a whole year’s worth of energy data is used.

“Another typical ploy is the use of partial PUE figures, a scenario in which the selective sampling of one area – say, the best-performing data room on-site – is promoted as being relevant for the entire facility,” he explains.

Meanwhile, Kevin Read, global infrastructure outsourcing UK senior delivery centre manager at IT services firm Capgemini, says there are lots of operators making claims about their PUE they simply can’t back up.

“I always find it amazing when organisations openly put items on their internet, showing their datacentre and – with a trained eye on the mechanical, electrical equipment on show – you can see masses of electricity being consumed by non-IT functions,” says Read.

“In which case, how is it they seem to make these amazing statements about their PUE when non-IT functions are consuming power?” he says.

As an example, Read cites firms touting a PUE score around the 1.1 mark, which he claims is difficult to achieve using mechanical cooling techniques.

“If the datacentres aren’t using fresh air cooling, and 99.9% of the datacentres in Europe do not, they couldn’t possibly be that energy efficient,” he says.

Despite the industry’s relentless pursuit of PUE scores that are as close to 1 as possible, it is worth noting that a facility with a score of two or more may not be as inefficient as it appears.

“PUE varies by climate, season and load. A facility that reports a PUE of 2.5 might sound bad, but could easily be a very effective design of 1.3 that is only loaded to 20% capacity because it’s newly constructed,” adds Ian Bitterlin, consulting engineer and visiting professor at the University of Leeds.

The true value of PUE

Pete Hopton, founder and chief visionary officer at datacentre liquid cooling specialist Iceotope, describes PUE as a “flawed metric” because some of the steps operators take to drive it down could have environmental drawbacks that may not be readily apparent elsewhere.

For example, in a bid to reduce their PUE scores, some facilities have adopted evaporative cooling systems, which require large amounts of clean water instead.

The latter is something that’s already monitored using the water usage effectiveness (WUE) metric, which measures how much H₂O a datacentre uses for cooling and other purposes.

It is an excellent metric that has been abused by marketing folks who publicise their PUE, which is strictly against the rules

Ian Bitterlin, University of Leeds

“A lot of datacentres have moved the PUE down from 1.7 by evaporating water and are now getting 1.1. Sometimes they’re big datacentres sited by rivers, but the amount of water some of them are consuming is immense,” claims Hopton.

“There is one datacentre that publishes its WUE data online and consumes around 20,000 litres of water per house. If you extrapolate that across its entire estate, you’ll find its water usage is probably more than the city of Portland.”

Licensed for use

Currently there is no way of telling if a datacentre operator is bluffing users with its PUE score, as it's not necessary to have the rating verified by a third party before publishing it, prompting many to question the metric’s value.

But, why should they need to be? The PUE metric was never designed as a means of comparing two competing facilities, but rather as an in-house benchmarking tool for datacentre firms looking to refine their operations.

As long as the provider carries out the calculations for its sites in a consistent way so the results are truly comparable, they shouldn’t necessarily need checking by an independent body.

“PUE should only be used internally to record improvements of a single facility against itself,” says Bitterlin, and should not be featured front and centre of a provider’s marketing materials.

“It is an excellent metric that has been abused by marketing folks who publicise their PUE, which is strictly against the rules,” he says. 

Clamping down

With the apparently wide misuse of PUE in the datacentre industry, is it too late to correct?

The Green Grid Association certainly doesn’t think so, having overseen the introduction of the metric in 2007, and is currently in the throes of getting PUE recognised as an ISO by the International Organisation for Standardisation. 

Transparency on PUE and adhering to the spirit and intent of the Green Grid’s PUE metric is the only way to generate real efficiency savings that translate to lower costs and carbon emissions for customers

Jason Liggins, Ark Data Centres 

Doing so should tighten up the rules around how PUE should be defined, measured, reported and used, as well as provide a comprehensive overview of how the metric directly relates to how a facility is operated.

Andre Rouyer, Europe liaison work group chair of the Green Grid, says the industry association anticipates PUE will be made a standard by early 2016.

However, operators will not be under compulsory pressure to follow the rules, but being granted ISO status could pave the way for certifications later down the line that show the claims an operator is making about their facility and its energy efficiency hold true.

“Official standards are not compulsory, but will become so if PUE is mentioned in a specification or government regulation,” Rouyer explains.

“A standard is not a regulation and is also not a certification, but a reference to use when regulations or certifications are in place to allow comparison and avoid misuse,” he adds.

Even so, CapGemini’s Read says if certifications were to follow, it would bring some much needed clarity to the industry, which can only be a good thing for users.

“Having a certification means it would be independently checked and if there is an ISO in place, you’d hope it’s going to be correctly audited, rather than companies simply making unfounded statements,” he adds.

This is a view echoed by Liggins, who says, ISO or not, it’s time the datacentre industry started telling the truth about PUE.

“When it comes to measuring efficiency and reporting PUE claims, it’s time for the industry to demonstrate complete integrity rather than massaging the figures to boost their sales pipelines.

“Transparency on PUE and adhering to the spirit and intent of the Green Grid’s PUE metric is the only way to generate real efficiency savings that translate to lower costs and carbon emissions for customers,” Liggins says.