Why Virtualization Worsens Your PUE

This audio was created using Microsoft Azure Speech Services

Virtualization is often touted as a “green” technology and with good reason. Consolidating lots of under-utilized servers into a handful of more fully loaded servers can save lots of electricity. But data center operators may be surprised to find out that, even though their electric bill goes down following a virtualization project, their power usage effectiveness (PUE) rating will likely worsen.

PUE is designed to measure the efficiency of the data center physical infrastructure, mainly power and cooling, in supporting the IT load.  It shows how much of the electrical power invested in that infrastructure is actually being used to power the IT load, as opposed to – well, something else.  A PUE of 1.0 is considered nirvana, as it shows that all the energy invested in the data center infrastructure goes to powering  the IT load.

Data centers in the real world, of course, won’t reach nirvana principally because they have what’s known as “fixed losses,” which is power that gets consumed by the infrastructure devices regardless of what the IT load is. Some of the big guys out there claim to get amazingly close to 1.0 though.  As the load shrinks post-virtualization, those fixed losses will increase as a proportion of the total energy consumed in the facility – that is, you’ve now got extra power and cooling capacity beyond what the IT load demands. As a result, your PUE goes up (i.e., worsens)..the numerator (total facility power consumption) remains the same while the denominator (total IT power consumption) decreases.

Reducing the PUE requires optimizing the power and cooling system such that it better aligns with the new, lighter load. You’ve got a few options for how to accomplish that, depending on your situation.

The option that’s likely to be most effective is to scale down overall power and cooling capacity. That may not be feasible if your power and cooling design isn’t modular, and thus not easily divisible. Or, if your infrastructure investment is relatively new or is shared with systems outside the data center (e.g., a chiller plant), replacing it or making significant modifications  to reduce its capacity may be impractical.

In such cases, options such as these may be more feasible:

  • Install blanking panels to reduce air mixing – especially important if you’ve got numerous empty rack slots because you got rid of lots of servers
  • Orienting racks into separate hot and cold aisles
  • Install air containment technologies to prevent hot air from mixing with cool
  • Adjusting fan speeds on cooling units to increase or decrease as needed depending on heat load
  • Turning off some cooling units
  • Removing unneeded UPS power modules for scalable UPSs
  • Align your temperature set points to the latest ASHRAE standards and use your cooling system’s economizer mode more often

By optimizing your power and cooling systems in these ways, you’ll maximize the benefits of virtualization.  Your PUE will not only be maintained or lowered to pre-virtualization levels, but your overall energy savings will be significantly extended as well.

To determine how much energy you may be able to save through a virtualization project, try the Schneider Electric Virtualization Energy Cost Calculator TradeOff Tool.  The interactive tool illustrates IT, physical infrastructure, and energy savings resulting from the virtualization of servers in a data center. You can input data regarding your data center capacity, load, number of servers, energy cost, and other data center elements to come up with estimated savings.

And to learn more about the effects virtualization projects have on data centers, check out APC by Schneider Electric white paper number 118, “Virtualization and Cloud Computing: Optimized Power, Cooling, and Management Maximizes Benefits.”

Tags: , , , , , , , ,