Average Power Density Vs. Peak Power Density

It might be a gross oversimplification but managing power density can be a real pain in the rack.  For those in the know, load balancing in the data center space is part art and part science.  On the latter end of the spectrum, did you know that there is much to be discussed about the difference between average rack power density and maximum rack power density across your data center floor?

It’s not unusual to be a little fuzzy on the differences between the two, but here it is at a high level:  average rack power density is power utilization density that’s averaged over a period of time and space, and peak power density is a maximum power reading at a specific point in time.

And yes, we may have been influenced by the late astrophysicist Stephen Hawking.  Read on.

First, let’s talk about some terminology. Because precision matters, it’s better to think of power density averaged over time as “spatial power density variation” and peak power density as “temporal power density variation.” With spatial power density variation, the average spatial power density is dependent upon the data center’s size and often is correlated with infrastructure capacities. 

Peak spatial power density depends on individual components within the system tied to specific design aspects.  With temporal power density variation, average temporal power density depends on regular application loads while the peak temporal power density depends on sporadic application loads.

Clear as quantum gravity?

We delve deeper into the topic of spatial vs. temporal power density variation in the white paper “Managing Variable Data Center Rack Densities,” where we also discuss real-life examples of peak allowable power loads and guide you through overprovisioning some or all racks to allow the overall data center to reach the allowable peak.

The paper ties in the density discussion with main data center goals of efficiency, capacity planning and uptime. It also provides guidance for using art and science to manage varying power equipment densities within data centers and shows you how to plan for handling the varying densities at the rack level.