Cooling-failure studies

A steady-state model tells you whether the room works with the cooling running. It can’t tell you what happens in the seconds after the cooling stops. Servers on UPS keep making heat, and the air in the room warms fast. Some racks start breathing their own exhaust within seconds.

A cooling-failure study follows the room through the event, second by second, and answers the questions a steady model can’t:

• How long until the first rack’s intake passes 27 °C, and 32 °C?
• Which racks get hot first, and for how long do they stay hot?
• Is the UPS ride-through long enough, and do the cooling fans need to be on UPS too?
• When the cooling comes back, how fast does the room recover? Does any rack keep climbing after the restart?

This is an engineering service, not a button in the software. CoolSim solves steady state. Our CFD engineers run the time-dependent study from your CoolSim model and hand you the answers.

Scenarios we’ve modeled

Since 2023 we’ve delivered 18 cooling-failure studies for data centers. They have covered:

• Total power loss, with IT load on UPS and every cooling unit stopped.
• Fans on UPS, compressors off: the air keeps moving but nothing cools it.
• Fans and pumps on UPS, riding through on the chilled water already in the loop.
• Staged restarts: fans back at 20 seconds, compressors at 80.
• Chillers back after one to two minutes, once the generators have picked up the load.
• Losing one unit: one fan wall of twelve, or three cooling units of eight.

An example: 30 seconds without cooling

We stopped every cooling unit in a 234-rack, 936 kW data hall for 30 seconds while the servers kept running, then brought the power back.

• The first rack passed 27 °C 14.5 seconds after the cooling stopped.
• The hottest moment came right at the restart. The hottest rack’s intake reached 33.5 °C half a second after the power came back.
• 17.5 seconds after the restart, every rack was under 27 °C again and stayed there.
• Without the restart, the same room kept heating. Every rack was over 32 °C by 84 seconds.

What you get

• Every rack’s intake temperature through the event, with the peak, when it happens, and how long each rack stays over 27 °C and 32 °C.
• Ride-through time: how long the room lasts before racks pass each limit, for each scenario.
• Recovery time once the cooling returns, rack by rack.
• Plots and floor maps of where the heat builds and where it clears first.
• Every assumption, written down. What is on UPS, when each piece of equipment restarts and how fast it ramps, and how air moves through a cooling unit whose fan has stopped. That last one can move individual racks by tens of seconds, so we state it rather than bury it.

What we need from you

• A CoolSim model of the room that solves in steady state. That is the starting point for every scenario.
• The failure sequence: what is on UPS, when the generators pick up, when fans, pumps, compressors and chillers restart, and how long each takes to reach full output.
• The IT load during the event, if it differs from the model.

How a study runs

1. You send the model and the sequence.
2. We agree the scenarios and the questions with you before anything is run.
3. Our engineers run each scenario and send you a report with the results above.
4. We walk you through the results.