Demystifying PUE Metric
Article originally published by 7x24 Exchange International -> Link to 7x24 Exchange Article
PUE has been a global accepted metric standard developed by The Green Grid, as a simple and quick way to benchmark the Data Center efficiency, and encourage continuous improvement.
Since the appearance of the PUE and its reciprocal DCiE, there has been a wide spread misconception about their applications and the real purpose, and scope of these metrics. This paper will try to clarify the PUE/DCiE metrics and their scope, in order to reduce any misconceptions, and also will try to lead the readers with a fresh view of PUE, and how it can be used to track Data Center energy efficiency.
Power vs Energy
The PUE original term was based on the power used by the IT equipment and the power used by the Facility systems/components that provide service to the IT equipment and the Computer Room. Power is typically measured in kilowatts units and is used during Design stages and can provide an immediate measurement in a point in time of the power drawn. The energy is the product of power (kW) consumed on a time base by any equipment, and the unit used to measure energy is kilowatt-hours (kWh). In the Figure 1 example, there is a graphical representation of power constantly drawn by some equipment (22kW) during three hours (time base). The resulted energy consumed by this equipment is 66kWh.
Figure 1
Power Vs. Energy
The reason for using energy instead of power is because it provides a more accurate picture of the status and the level of efficiency over a period of time. Power typically fluctuates and it can drive us to erroneous or inaccurate conclusions about the performance/efficiency of any equipment/system over time.
However power can help to give us an idea about the performance in a particular point in time, and helps to project scenarios over future system performance.
PUE Calculation
PUE can be calculated using either power units (kW) or energy units (kWh). PUE is the ratio of (Total “Data Center facility” Energy or Power) to (“IT Equipment” Energy or Power). As a convention “IT equipment” is a designation that includes Information Technology, Networking, Cybersecurity and Telecommunications equipment's. The “Data Center Facility” considers all the energy or power drawn by the IT Equipment and every Facility system/component that serves to the IT Equipment and the Computer Room, including the Cooling System, Electrical System, and every ancillary system with the purpose of serving to the Data Center’s mission.
Figure 2
PUE: Power and Energy versions
The PUE is basically a comparison of how much energy is consumed by the Data Center Facility in terms of “IT equipment units”. In other words it tells how many times IT Equipment energy is consumed by the Data Center Facility.
PUE resulted value is always > 1.0
Resulted PUE values varies from 1.0 to infinity. As the value approaches to 1.0 the better is the energy used by the systems/components of the Data Center Facility. In the Figure 3 there are some references of PUE values and their level of efficiency.
Figure 3
PUE World and U.S. accepted levels
Figure 4 exemplifies a Data Center Facility and their typical Facility systems/components with the IT Equipment they serve. It is easier con classify or aggregate the components by systems in order to discriminate the energy consumption by area of opportunity.
In order to understand PUE is important to differentiate Data Center facilities equipment, according to its mission or purpose, they can be classify in two main groups: IT Equipment and Site Infrastructure systems.
•IT Equipment accounts for every Information Technology and Communication Equipment serviced by the Data Center.
•Site Infrastructure is composed by all the Facility systems/components that provides a service to IT Equipment.
When we look at PUE basically we take a look of how efficient the Site Infrastructure that serves the IT Equipment is.
Figure 4
Simplify view of Data Center facility systems and components
As shown in the figure 5, PUE formula is graphically represented to give an idea of the systems involved.
Figure 5
The PUE formula graphically represented
In the figure 5 we can easily see that the “Total DC Facility energy” will be represented as “IT Equipment energy” multiples. The value resulted from PUE calculation would be “IT Equipment energy” units, in example a PUE=2.0 would be 2 “IT Equipment energy” units or 2 times the “IT Equipment energy”. The same happens with PUE=1.5 it would be 1.5 times “IT Equipment energy”. The figure 6 shows the representation of theses examples values.
Figure 6
PUE=2.0 vs PUE=1.5 Graphically represented
PUE gives an idea of magnitude about how much the Data Center Facility energy is by comparing it with the IT Equipment energy consumption. So it will show the Data Center Facility energy in terms of IT Equipment energy multiples. It makes a fair comparison, by contrasting the IT Equipment energy consumption against the Site Infrastructure energy consumption that was designed to provide different kind of services (Cooling, Power, Lighting, Security, etc) to the IT Equipment and to the Data Center’s mission. This approach has given to the Data Center industry a race to benchmark every Data Center, and showcasing technologies and design approaches that makes improvement to Data Centers with the PUE in the spotlight. However, PUE is really meant to be used for a Data Center to measure its performance improvements, not to compare different Data Centers.
How to use PUE as a way to improve Data Center energy consumption
PUE is a metric that can help not only to assess our current situation about the energy consumption but also to set future goals and easily follow up every improvement.
As a metric for Data Center Site Infrastructure it accounts about the energy consumption of:
•Mechanical/Cooling System (Chillers, CRAC/CRAH, Condensing Units, Humidifiers/Dehumidifiers, etc)
•Electrical System (UPS technology, Power Distribution, Generators, Transformers, etc)
•Security Systems (Fire Protection, EAC, Intruder Detection Systems, etc)
•Lighting system
•Ancillary Systems (Monitoring technologies, SCADA, etc.)
All of these facility infrastructure components are considered to be accounted by the PUE, and the related energy can be tracked independently with the approach shown in the next pages.
A way to easily see the information resulted from PUE is representing it graphically like the figure 7. In that figure PUE is initially decomposed by differentiating the part related to IT equipment (constant value) which always is 1, and the part related to Site Infrastructure (variable value) which is the area of opportunity for improvements or the optimization energy area.
Figure 7
Site Infrastructure as the Optimization Energy Area for PUE=2.0 and PUE=1.5
In the figure 7 can be noticed the PUE minimum theoretical value of 1 as the part related to IT equipment. In the real world PUE=1 never can be reached. PUE will be always > 1 and maybe the best it can get could be 1.0x and that would be a very exceptional number.
So the area of energy optimization will be the part of the PUE related to the Site Infrastructure trying to get the lesser possible number approaching to 1. That demonstrates that PUE is a metric designed as a KPI to target only Site Infrastructure energy performance. The part related to IT Equipment Energy is accounted as a constant value as a reference to contrast Site Infrastructure energy performance. We could say that IT Equipment Energy is treated as a “Black Box” for the PUE purposes. Therefore, PUE may hide inefficiencies in the IT Equipment.
Partial PUE (pPUE) of Site Infrastructure Systems/Components
Partial PUE can be used as a tool to make PUE a more precise tool in order to deep down in every systems of the Site Infrastructure to look for improvements. One of the uses of partial PUE is to measure one system that is part of the Site Infrastructure or it can go even further by decomposing a particular system. That can gives a better picture in order to understand where the energy is being drawn and how we can improve a component within a system.
Partial PUE can gives also a quick idea of how efficient a system/component is.
Figure 8
Partial PUE of System/Component
In the table 1 there is an example of Data Center facility energy consumption in 1 year.
Table 1
Example of a Data Center energy consumption in 1 year (Case Scenario)
In the next example (Figure 9) there will be partial PUE (pPUE) and PUE calculations from Table 1:
Figure 9
Partial PUE and PUE calculations (Case Scenario Table 1)
Decomposing PUE
One of the advantages of pPUE is the ability to decompose PUE when looking for areas of improvement. That can be done by resting pPUE of any system/component minus 1, then you get the part of the PUE related to this system/component. The figure 10 is a graphical representation example of how would be a decomposition of PUE using Table 1 data.
Figure 10
Example of PUE Decomposition technique (Case Scenario)
Table 2 shows more detail about the example of PUE decomposition. In Figures 11-A and 11-B, can be seen pie graphical views from table 2.
Table 2
Example of Data Center Decomposition (Case Scenario)
By decomposing the PUE, it is possible to understand how every system/component affects PUE. It can also be represented in percentage. Here there is a set of formulas that help in decomposing PUE.
Figure 12
PUE Decomposition Formulas
Figure 13
Site Infrastructure Energy Distribution Formulas
Figure 14
DCiE Formula
DCiE is the reciprocal of PUE, and is another way to look about Data Center energy efficiency. Basically DCiE is the portion of Total DC Facility Energy related to IT Equipment in percentage.
DCiE is a metric that can be used independently of PUE, however by using these metrics together can give a more useful picture of Data Center Energy Consumption.
Extrapolating PUE
Another advantage of PUE when we look for energy efficiency improvements in terms of absolute energy consumption, is the easy way to extrapolate the resulted energy efficiency of any Data Center.
PUE can be used to target a point of efficiency and comparing it with the current PUE value, in order to see how much energy can be saved in kWh and also in financial terms.
That is one of the most important aspects of PUE. As we treat IT Equipment energy as a constant value, we can effectively project Site Infrastructure improvements by estimating future PUE to later calculate the savings.
However it´s important to be careful in the way the improvements are made in order to not affect negatively the IT Equipment intake conditions (temp/cfm). By affecting negatively IT equipment intake conditions, might have an important increase in IT Equipment Energy consumption causing a distortion of the IT Equipment energy (increased consumption) that will create a false improvement of PUE.
The figure 15 is an example of savings by extrapolating PUE.
Figure 15
Example of Extrapolating PUE
In the Figure 15 example there is a saving of 4,462,350 kWh provided by reducing the PUE value from PUE = 2.0 to a future target value of PUE = 1.5.
If the utility rate would be for example of US$ 0,013 per kWh then the resulted savings would be US$ 58,010 in a year.
Those savings can be used to be applied to Site Infrastructure ROI, in order to justify the improvements.
Conclusion
It can be affirmed without any mistake that PUE is a metric related to Data Center Site Infrastructure. The more effort it is put on optimizing the Site Infrastructure in terms of energy efficiency the more impact will be reflected on PUE. PUE is arguably the best quick way to get a first impression or initial idea about how energy efficient a Data Center is. Additionally if it is used in conjunction with partial PUE System/Component (pPUE) and PUE decomposition techniques, it can provide a more deep insight about what components consume more energy and therefore where the improvements should be made in order to generate better results.
By tracking the PUE decomposition to keep a historical record , would give to Data Center operators and managers with a powerful tool to follow up any improvement, and could also be a tool to continuously monitor the energy performance of any system and component.
Author: Jorge A. Gil - jagil@seresdc.com