Power Usage Effectiveness (PUE) is a metric created by The Green Grid to determine the energy efficiency of a data center. PUE is a ratio, with overall efficiency improving as the quotient decreases toward unity. In other words, PUE measures the extra power needed to power, cool, protect, and manage the IT load in a data center.
Since the issuance of PUE, many data center managers are under pressure to drive the PUE of their facility lower. The notion of driving energy efficiency is a noble one, but the drive toward a PUE target is not a pragmatic endeavor.
PUE should be viewed as a tool, not a destination. While the PUE is based upon very simple math, PUE comparisons are very difficult. In fact, it’s dangerous to compare PUE values between data centers because so much depends upon where (with respect to the facility infrastructure) PUE readings are taken, as well as which data center energy components are included in the scoring. There are simply too many variations from facility to facility and from operator to operator to know if one has a true “apples to apples” comparison.
Enterprise executives concerned with this metric often reference the industry innovation giants of Google and Microsoft, who claim PUE values below 1.25. Companies such as this have very specific applications and architectures that enable them to take energy efficiency steps that may not be appropriate for the general enterprise. The strategies that some companies are using to achieve a certain PUE are extreme and are rarely suitable or fundable for the typical enterprise data center.
The enterprise data center operator should be aware of cause and effect. If the data center is located in a shared-use building, PUE will be impacted by the non-data center power draw. High availability architectures are the nemesis of energy efficiency. Redundancy worsens PUE. Virtualization and consolidation often have a counter intuitive impact on energy efficiency. The electric bill may be better because of optimized use of server capacity, but PUE may be worse unless power and cooling are simultaneously downsized to align with the IT load. Similarly, raising server inlet temperature may improve PUE because of higher efficiency operation of cooling systems, but the electric bill may be worse if increased server fan power exceeds cooling system savings.
The data center operator should first understand PUE, then choose a measurement strategy, carefully consider options for PUE improvement, and make sure you don’t have a “PUE ending.”