Power usage effectiveness is a common way to track data center power, but to adequately minimize power use, admins...
need more information than what traditional temperature probes and power strips can provide.
Power usage effectiveness (PUE) is a metric that compares how much energy a data center facility uses versus the total IT equipment energy consumption, and it offers a general picture of energy use. Organizations can reduce data center power usage, but without upgrading the cooling system, PUE can still worsen.
Admins can implement data center infrastructure management (DCIM) software to calculate PUE more accurately and optimize data center efficiency.
How to calculate PUE and PI
To calculate PUE, admins should divide their total facility energy by the total IT equipment energy. Admins can calculate total facility energy with a utility meter or bill. They should then determine the total IT equipment energy with remote power panels or smart rack power distribution units (PDUs).
The Green Grid regularly updates the PUE standard to include more data and account for network operations, power interruptions, multiple data centers and power consumption over time. This means there are four levels of PUE: PUE0, PUE1, PUE2 and PUE3.
Each PUE level uses different methods to calculate the total IT equipment energy. As the levels increase, they provide a more precise PUE measurement.
PUE0 measures power for one data center in one particular instance, and admins can use the uninterrupted power supply (UPS) output to define the IT energy load. PUE1 also accounts for UPS measurements, but admins must convert all the readings to kilowatts per hour and perform measurements monthly.
With PUE2, admins calculate IT energy with PDU data -- instead of UPS metrics -- and measure it in kilowatts per hour. PUE3 requires admins to source the IT load data directly from the hardware and continuously gather information in intervals of 15 minutes or less.
Depending on how accurate admins need the PUE measurement to be, they can use the guidelines from PUE1, PUE2 or PUE3 to calculate the IT load. Admins should always aim for a PUE calculation of 1.0.
Admins can also measure cooling effectiveness with the performance indicator (PI) benchmark, which accounts for efficiency, performance and resiliency, and how the metrics affect one another. To calculate PI, admins can use the PUE ratio, IT thermal conformance and IT thermal resilience.
How to manage data center power usage with DCIM
DCIM software combines processing demand -- how much hardware workloads need to properly run --- with infrastructure capacity -- how much data the servers can actually handle. With the increased transparency, admins can match workloads to the appropriate amount of servers and minimize IT equipment power consumption.
Robust DCIM software can gather watts per hour, gaps between planned and actual energy usage, CPU utilization, internal data center temperatures, and available cooling resources. Then, it can display these metrics on a unified dashboard and provide a more accurate picture of data center power usage.
Certain software gathers data, calculates PUE and includes the metric as part of the dashboard. Market offerings with this capability include Nlyte Energy Optimizer, Panduit SmartZone, Sunbird Power IQ, Raritan Power IQ DCIM and Schneider Electric StruxureWare.
DCIM software also tracks server workload and indicates which servers or virtual machines admins can actually shut down if the workload goes to zero or below an admin-defined threshold. With server workload information, admins can identify reduced server usage, reconfigure loads to use the minimum number of servers and shut off any unnecessary hardware.
Over the long term, the reporting features of DCIM software provide data on which data center subsystems and racks consume the most electricity and cooling energy. DCIM tools give admins the ability to set actions during certain times of day -- such as workload calculation, data collection, software patching and provisioning -- to automate server software and provide a clear picture of the total power system, including power chain dependencies and data center power usage.
Admins should know which applications and hardware will go offline during a power failure to assess and mitigate vulnerabilities, as well as to speed up recovery from potential failures. DCIM software can indicate if admins have enough capacity in each part of the power chain to install or relocate hardware without increasing operational risk.