4 Tips for Keeping Your IT Equipment Cool When the Power Goes Out

In my last post, I shared some of the ways that data center energy efficiency efforts can cause temperatures to quickly rise in a data center following a power outage. That was the bad news.

The good news is there are ways to counteract these dramatic temperature increases. They are outlined in great detail in white paper 179, “Data Center Temperature Rise During a Cooling System Outage,” but I will share some highlights here.

First, realize that you’re in a disaster situation, which means making some decisions on which resources you really need. Powering down all but the most critical resources by itself will help keep the temperature down.

Otherwise, the paper outlines four strategies to slow the rate of heating:

  1. Maintain adequate reserve cooling capacity
  2. Connect cooling equipment to backup power
  3. Use equipment with shorter restart times
  4. Use thermal storage to ride out chiller-restart time

1. Maintain adequate reserve cooling capacity

The trend toward “right-sizing” cooling capacity such that it closely matches the IT load makes sense from an energy efficiency perspective, and works perfectly well under normal conditions. But having little or no excess cooling capacity means it’ll take a good long time to get the data center down to a normal temperature after a power outage.

The solution is to scale the bulk cooling, such as chillers, and cooling distribution system (CRAH units) as the IT load increases. This will provide some reserve cooling without sacrificing data center efficiency. Maybe you’ve got a data center designed to handle a maximum IT load of 1MW but initially the load is only 100kW. Your chilled water plant piping will be sized for the maximum load but you can install chillers sized for maybe 140kW of IT load. That’ll give you some spare capacity in the event of a power outage.

2. Connect cooling equipment to backup power

Failure to connect CRAH fans and chilled water pumps to a UPS can result in dramatic temperature spikes following a power outage – easily 9°F/minute (5°C/minute) or more depending on density and room layout.

Placing CRAH or CRAC fans on UPS will help by limiting recirculation of heat from the IT exhaust to the IT inlet, instead helping to transfer it to the data center at large. Even more effective is to also place pumps on UPS, especially in systems that use chilled water CRAH units. The chilled water and piping system can significantly extend the window of operation following a power failure, the white paper says.

3. Use equipment with shorter restart times

When chillers lose power for more than a few milliseconds, they typically have to restart, which traditionally took 10 to 15 minutes. But some newer chillers restart much faster, in as little as 4-5 minutes. This is important following the initially power outage but also when power is eventually transferred from the generator back to utility power. That typically involves a momentary brownout, causing another chiller restart. While quicker restarts alone may not prevent an unacceptable temperature increase, in lower density data centers it may be all you need to weather the outage. The quick start chillers also enable you to keep chilled water and IT temperatures at higher levels during normal operating conditions with less risk of exceeding acceptable temperatures under emergency conditions.

4. Use thermal storage to ride out chiller-restart time

If your data center uses a chilled water cooling system, installing an additional chilled water storage tank may be enough to get you through the outage. In fact, if your chilled water pumps and CRAH fans are on UPS, you’ll likely see very little departure from normal operating temperatures.

Low-pressure thermal storage tanks are relatively inexpensive, as they can be made of plastic. They are recommended especially for high-density data centers where even a short period without cooling can create problems.

For more details on how to make sure your data center stays cool during the next power outage, check out white paper 179, “Data Center Temperature Rise During a Cooling System Outage.”

Tags: , , , , , ,


  • Actually, the best way is to place the *entire* cooling infrastructure (fans, pumps, chillers, cooling towers, etc.) on UPS and generator. (Yes, I had an entire 3-story building where we had done this! There was not a single circuit not on UPS.)
    Unfortunately, this is costly and almost never done, especially when the UPS is battery-based. But it prevents the chillers from dropping offline during the ATS transfers to/from emergency power.

    Some DRUPS systems, due to their “no break” abilities, can also be configured like this.

Comments are closed.