In the world of data centres, Power Usage Effectiveness (PUE) has long been touted as a metric for measuring efficiency. However, it’s time to challenge this conventional wisdom and unravel the real complexities surrounding PUE.
Let’s take a fresh look at the data centre landscape and explore the true factors affecting energy consumption.
The Limitations of PUE
Contrary to popular belief, PUE is not an all-encompassing measure of data centre efficiency. Using it as a proxy in this way is simply inappropriate and can distract us from addressing the actual issue of data centre power consumption.
Instead of falling into the PUE trap, we need to shift our focus towards more meaningful solutions that can help improve overall data centre energy efficiency.
A Shift in Perspective
Let us set the record straight: A data centre does not consume power; it merely adds an overhead to the energy consumed by the digital infrastructure it houses.
One goal should certainly be to minimise this overhead to the greatest extent possible – precisely what PUE aims to measure. However, this is only part of the overall picture.
While we can improve PUE by increasing the IT load relative to the building load, it doesn’t address the bigger picture of overall energy efficiency. In some situations, it is possible to make the PUE value smaller (theoretically ‘better’), and yet increase the number of kW/h consumed, something that would not normally be considered to be an efficiency gain.
The True Culprit: The IT Load
To achieve substantial energy efficiency gains, we must turn our attention to the heart of the matter – the IT load hosted within data centres.
It’s no secret that low server utilisation, overprovisioning of capacity, outdated servers, and energy-inefficient applications are common challenges in data centres and yet they can make PUE numbers look better. Removing these issues and consuming less energy can make PUE look worse.
Instead of chasing increasingly elusive improvements in PUE, we should prioritise reducing energy consumption at the IT level, which will naturally lead to a proportional reduction in building overhead (PUE).
Consistency and Recognition
One of the glaring issues with PUE lies in its inconsistent measurement and reporting practices.
Many references to PUE lack awareness that it should be calculated and reported using standardised KPIs from the International Standards Organisation (ISO), specifically ISO/IEC 30134-2, which among other things requires PUE to be an annualised average, not a point in time taken during the coldest part of the year. Anything else is simply not true PUE and design PUE, Partial PUE or point measurements should not be used to represent true PUE.
Interestingly, this same standards series includes metrics for Water Usage (WUE), Carbon Usage (CUE), Energy Re-use (ERF), and Renewable Energy Use (REF) – untapped opportunities for benchmarking and standardised reporting that the sector has been slow to recognise and utilise.
Embracing the Real Opportunities
Let’s veer away from the PUE fixation and embrace the real opportunities for energy efficiency within any data centre.
By focusing on the IT stack rather than solely blaming the building, we can unlock the true potential for reducing energy consumption and improving environmental sustainability.
It’s time to re-evaluate our approach to data centre efficiency. PUE might have been a stepping-stone, but we must recognise its limitations and shift our focus towards the IT load as the primary driver of energy consumption.
By acknowledging the importance of standardised measurements and exploring untapped metrics, we can navigate the path to a far more sustainable and energy-efficient future.