Skip to main content

Computational Fluid Dynamics

By September 2, 2014December 26th, 2021Briefs

Data servers are like ice cubes on a hot day. When you treat them right, they provide wonderful, life enhancing services – like a margarita, or in the case of a server, Facebook! But they require a gentle hand. If they get too hot, bad things happen. So providing sufficient cooling is very, very important.

Often, server rooms are overcooled for just this reason. However, servers aren’t like people in that they don’t complain when the temperatures get above 72F (no offense, Mom!). In fact, today’s recommendations for server inlet temperatures can actually be closer to 80F, and they’re happy as a clam.

But just because the air in the server room is 80F doesn’t mean the air entering the server is 80F. In server rooms, one can’t assume that all the air flows perfectly from the air inlet to the air outlet. Air circulation patterns can do unexpected things at the small scale.

For instance, without proper flow control, air can come out the back of the server and immediately be turned around and sucked back into the front, which can cause the inlet temperature to reach dangerously high levels.

This problem has typically forced designers and operators to apply wasteful solutions and simply turn the temperature down in the entire room. To avoid this, designers and operators need a more refined tool to understand the small-scale airflow patterns around the individual servers.

CFD to the rescue

Computational Fluid Dynamics (CFD) software gives engineers and owners the understanding of how air flows throughout the room as well as around the servers themselves. This understanding can then be used to better optimize the systems that condition the air in the space. For instance, sometimes warm air is recirculated back to the server inlet. In this case, rather than cooling down the entire room to lower the server inlet temperature, just stop the recirculation!

CFD can predict the air patterns in a room such as a server room. This shows where hot air and cold air are moving and can help predict “hot spots”, which are locations where the air is too warm for the servers. By identifying these spots, engineers and facilities personnel can put up barriers to reroute flow where it needs to go.

The effect of this is that the room can be operated at warmer temperatures. Warmer temperatures mean less cooling is necessary and systems operate at higher efficiency. It can also greatly expand the period over which free cooling with outside air is sufficient.  Because data centers typically operate in a cooling mode for a majority of the year, this can lead to huge savings that wouldn’t have been possible without knowing how the air actually moves through the facility.

Michaels Energy

Author Michaels Energy

More posts by Michaels Energy