by Arthur Cole, IT Business Edge Arthur Cole spoke with Bob Hunter, CEO of Trendpoint. Critics of the Power Usage Effectiveness (PUE) benchmark are quick to point out that it does not necessarily provide a complete picture of overall efficiency. By focusing on overall consumption, the standard does little to point out particular trouble spots within the infrastructure. But Trendpoint has devised what it calls the Micro PUE, which seeks to gauge energy use at each individual cooling unit. Hunter argues that this provides more useful intelligence about what is really going on in the data center, and what can be done to improve energy efficiency. Cole: What is the Micro PUE approach, and how does it differ from the standard PUE measurement?
“ We measure the PUE from the perspective of each individual cooling unit, effectively slicing the whole PUE pie into coherent slices that correspond to individual cooling-unit-to-IT-load segments.”
- Bob Hunter, CEO of Trendpoint
Hunter: Micro PUE is a management-centric approach to PUE management. PUE is a fine benchmark, but suffers from two problems as an actionable management tool. First, it is a single-point-in-time measurement. That is, it is akin to reporting your bank account balance at any given point in time and is therefore subject to allowing the user to pick and choose the most appealing figure. Second, it is a macro-figure that includes all IT cabinets and all cooling systems, thus, one has no idea where the individual problems exist that can keep PUE high. Micro PUE is based on the fact that PUE is a cooling-centric number. That is, it is only possible to lower PUE by lowering the cooling energy vis-à-vis IT loading. With this in mind, we measure the PUE from the perspective of each individual cooling unit, effectively slicing the whole PUE pie into coherent slices that correspond to individual cooling-unit-to-IT-load segments. With this data, users can for the first time see where their problem areas are in controlling their PUE and, equally valuable, provide trending of this data to see how IT load changes and cooling infrastructure changes are affecting PUE. Cole: It still seems like a very finite benchmark. What other metrics would be needed to provide a clearer picture of overall data center efficiency? Hunter: The Micro PUE segments all work together to provide a complete view of PUE. For example, in a data center with 10 air handler units, the user would be able to view each of the 10 PUE “slices” from least efficient PUE to most efficient PUE. They would then be able to change loading profiles and infrastructure – for example, changing floor tiles, adjusting hot isles or cold isles – and watch the change in Micro PUE for that problem segment in real time. Thus, you have a feedback loop that allows users to both manage their PUE in real time by segment and to keep track of long-term changes in PUE. Cole: Is there a risk that energy efficiency is such a complicated thing to measure that any attempt will allow organizations to game the results, making the entire process ineffectual? Hunter: The problem of gaming the results is already a very real issue. That is our point about the bank account analogy. Users can simply pick and choose a PUE number on the best minute of the best day of the year. On the other hand, we suggest using actual energy data in the same billing units used by the utility — kiloWatt hours. By using kWh for minimum time periods such as one day, one week or one month, you get the average PUE values over that time period. That would avoid the ability to game results and put all users on an equal footing. In the same way, by allowing users to see Micro PUE in both real time for active management changes of IT and cooling infrastructure as well as longer timer kWh over any chosen time period, a very true picture emerges of which segments are helping and which micro-segments are hurting overall PUE. View article on IT Business Edge