Tuesday, 25 September 2012

Damned Statistics...

Do you know your PUE from your FVER? Energy metrics are now regarded as a key means of demonstrating your 'greenness'. But which system should you be considering?

Popular efficiency metrics such as PUE (power usage effectiveness) have undoubtedly been successful in getting the data centre industry to wise up to the problem of energy efficiency. However, while many organisations have followed its principles with rigour, its popularity has led some in the industry to 'juke' the stats, rendering hollow figures, rather than useful efficiency measurements.

This was one of Greenpeace's main concerns in its 'How green is your cloud report' (featured in Green IT), as businesses were able to provide low PUE figures, it stated, without actually being energy efficient or environmentally responsible.

 PUE is by no means the only efficiency metric within the data centre industry, of course, but no other has achieved the same level of adoption industry-wide. The new data centre energy efficiency metric FVER (Fixed to Variable Energy Ratio) aims to change all that. Rather than measuring IT in the same way you would utility services such as water and electricity, FVER targets waste. It might serve the industry well in the future to incorporate measures such as this as part of a wider efficiency strategy within data centres - and to use any metric as only one factor in efficiency measurement, rather than the be-all and end-all statistic. "It's on the basis of PUE's success that we can now evaluate the potential of new efficiency metric FVER to help improve current efficiency measures and instil necessary behavioural changes within the industry," asserts Peter Hopton, Iceotope CTO and founder.

"Given that FVER is measured as a determinate figure from one to infinity, like PUE the lower the number, the better. This makes it likely to be understood and embraced by C level executives, as well as data centre operators."

 The premise behind FVER is simple; it assumes your data centre is made up of a sum of two loads, a fixed load that would exist, if the data centre was inactive, and a variable load that would be maxed out when the data centre was full to capacity. This metric allows data centre operators to measure the difference between the two states and, as a result, evaluate the efficiency of their systems.

"In crude terms, the greater the difference between the fixed and variable energy figures, the lower the FVER figure," Hopton explains "But, crucially, the fixed energy figure has a significantly larger impact on the solution than the variable energy figure. As a result, this metric encourages data centre operators to target the fixed load, rather than the variable energy, and following this method could lead to data centres operating much more efficiently.

To put forward an example, if your data centre's average server utilisation is ten per cent and you are given two options - reduce the fixed load or reduce the peak power consumption, while keeping the fixed load constant - which would you choose? Reducing the fixed load will be nine times more effective than reducing the peak power consumption. FVER is reflective of this, and makes both energy waste and potential efficiency gains much more visible." With technology in general, functional value is largely consistent with energy/ resources used or financial expenditure. "For example, you'd expect your television to require more energy when transmitting video than when on standby, your car to use more fuel when in motion than when stationary and to be charged more by your mobile provider for your current model than the long forgotten handset locked away in a desk drawer," he points out. The data centre industry does not follow this system, as a rule. Energy used remains at a somewhat consistent level, regardless of the 'useful work' being conducted - this ensures costs also remain high when minimal work is actually being done. "As puzzling as this may sound, what really baffles is that these facilities may not even be considered inefficient," adds Hopton. "In fact, dependent on the metrics and measurements used, they may be seen as running highly efficiently." Like all metrics, FVER is not perfect. Systems such as liquid cooling will have a worse FVER rating than an air cooled system, despite being much more efficient - as fans are a strong (but wasteful) variable load. "FVER will, however, cause people to think about the fans inside their servers, encouraging them to throttle down. FVER also targets software, encouraging it to allow the server to reduce its power consumption when the software is underutilised."

Ultimately, FVER should be used as part of a larger sustainability structure, utilised alongside other measures and efficiency metrics like PUE and combined with straightforward common sense, Hopton advises. "Whether there could be an efficiency metric that covers all areas of energy efficiency, or indeed whether there should be, is a debate for another time," he suggests. "If FVER can somehow help users recognise where their systems are wasteful or inefficient, this could pave the way for substantial energy efficiency measures in the future, similar to those we first saw with the introduction of PUE."

BCS data centre specialist group secretary and Romonet CTO Liam Newcombe is another advocate for FVER. "A major part of the potential efficiency improvement in the data centre is now locked up in the fixed 'base-load' power draw. Correcting this issue should be the next priority for data centre operators and the wider industry," he argues. "In an ideal data centre, whatever the power draw, when the IT platforms are at peak load, it would be zero when the IT platforms are delivering no services. The DC FVER metric provides a way for operators to measure how well their IT and site energy consumption tracks the useful work delivered by their IT platforms." Peter Hopton, Iceotope CTO and founder.