Since the power usage effectiveness (PUE) rating was created in 2006, a number of people have tried to improve on it. But the latest effort has stirred up some controversy by trying to extend it to include water use.
PUE was created to help data centers reduce their carbon footprint by cutting their consumption of electricity, which is normally generated from fossil fuels. To help reduce this, PUE offers an indication of how efficient a data center is at delivering electrical energy to the IT equipment it contains. It has a simple formula: the total energy used by a facility, divided by the energy that is delivered to the IT racks.
Obviously enough, any energy used outside the IT kit will increase the top line of the fraction and increase the PUE, taking it further away from the ideal theoretical figure of 1.0. So achieving a better PUE has become a matter of choosing less energy-hungry infrastructure for tasks like cooling and power distribution.
But there is more to efficiency than electricity usage, say members of The Green Grid – the non-profit industry group that promotes PUE. The group created a separate metric for water usage effectiveness (WUE), which is simply the ratio of water used to IT power.
WUE has not achieved any real traction in the market, but water usage has become a concern, particularly in parts of the world where water is expensive, or which suffer from droughts.
Meanwhile, concentrating on PUE could actually be pushing data centers to consume even more water, according to a recent blog post by Jack Pouchet of Emerson Network Power: “Although it was clearly launched as a tool to be used wholly within the confines of a single facility, PUE has nonetheless become the de facto universal metric for comparison shopping when evaluating data centers. This behavior has led to numerous unintended consequences, not least of which is the staggering increase in the use of water by data centers.”
Data centers have been moving away from mechanical cooling systems, which use a lot of electricity and comparatively little water, to outside air cooling systems, which are often boosted by evaporative cooling.
The move is only possible in countries where clean water is readily available, and where the climate allows cooling by evaporation, but the effect on PUE has been remarkable. Data centers that use outside cooling have seen their PUE drop from around 2.0 to 1.2 or less.
The effect on PUE is quite legitimate, because evaporative cooling is a natural process driven by the difference in vapor pressure between the liquid and the air around it. It does not need electricity to drive it, and it is not responsible for carbon emissions. Moving to evaporative cooling cuts the carbon emissions associated with a data center.
What about the water use? Pouchet’s suggestion for “PUE revised” (PUER) adds in a factor for water usage, the water equivalent energy (WEE), and would thus increase the PUE of data centers that use a lot of water.
Level playing field?
There is some justification for this, and Pouchet’s idea has met with enthusiasm from some who see it “levelling the playing field” for data centers in warm and humid climates, where evaporative cooling won’t work, or in areas where water is in short supply and the method is unaffordable. PUE is a metric created by the Northern Hemisphere, for the Northern Hemisphere, say the critics.
If PUER caught on, it would certainly have a significant effect on efficiency figures.
But the WEE factor has been controversial. Pouchet uses the energy required to evaporate the water, which is 2,257kJ (0.627kWh) per liter. While it is true that this energy is “used” in evaporating the water, it has come from available low-grade heat and has not incurred any electricity usage or carbon emissions.
Robert Tozer of Operational Intelligence describes this as a “thermodynamic flaw” in PUER, and others have levelled similar criticisms of false energy accounting. Essentially, it takes an item that doesn’t feature on the energy balance sheet and adds it to the debit side.
Mixing water and power
If PUER caught on, it would certainly have a significant effect on efficiency figures. A 1MW data center with evaporative cooling might have a PUE of 1.1, but it would have a PUER of 1.4. If it uses indirect evaporative cooling, its PUER could go up above 2.0.
But will it catch on? It will be put to The Green Grid, where Pouchet has been a long-term participant, for evaluation: “We are moving forward with this via The Green Grid processes,” he says.
But PUE has become well established over nearly 10 years, and has even reached the stage of being published as a draft international standard (DIS 30134-2) by ISO – the International Organization for Standardization. Adding the WEE factor would make the measure more complex and undo the work of creating that international standard.
Moreover, Ian Bitterlin of Critical Facilities Consulting (p28), says that water is already included in PUE. Behind the simple PUE formula, The Green Grid and other bodies such as ASHRAE have published a lot of information on how to calculate all the equivalent energy for parts of the data center, including the water used in an evaporative cooling system. “Read both V2 of TGG and ISO 30134-2 and you will find the proxy kWh/ unit tables,” says Bitterlin. “The energy for evaporation or pressurisation for adiabatic spraying has always been a part of the facility overhead power.”
The bottom line is that Northern Hemisphere data centers do indeed get a better PUE figure than those in other climates. However, it is an unavoidable fact that they also have a lower carbon footprint.
Water and power are connected in a complex way. Mixing the two runs the risk of clouding the issue.