Modeling IT Containment
This webinar shows how to represent IT containment around data center racks in a CoolSim computational fluid dynamics (CFD) model. It is aimed at engineers who build or review data hall models and need to decide how much leakage and detail to include.
What the webinar covers
The talk starts with the physics that sizes the airflow. Heat removed equals airflow times air density times specific heat times the temperature difference, and the same relationship governs both the servers and the cooling units. Server airflow per kilowatt has fallen over the years as the temperature rise across servers has increased. The presenter shows the expected-temperature-rise input for racks: about 123 cubic feet per minute (CFM) per kilowatt at a 25 °F rise, and about 88 at 35 °F.
The example room is a hot aisle contained, non-raised-floor design with air handlers outside the room and a ceiling plenum return. The pre-simulation report compares the air the cooling units can deliver with the air the racks demand.
Three ways to model the containment are then compared on the same room:
- Baffles on all sides, with the percent open area set to represent leakage.
- Racks sized slightly smaller than the real cabinet, with small gaps around them and solid blocks above, to represent leakage at the rack rails and under the rack.
- Detailed racks built with the rack builder, with individual servers or a full cabinet with rails, doors and gaps.
A recorded screencast shows the build: dragging in a baffle, aligning edges, closing gaps, grouping, and cloning a group with an offset to lay out a row of containment. The comparison covers mesh size, run time, leakage, room pressure, temperature slices and the airflow through the ceiling grilles. The session ends with audience questions on making objects transparent and on modeling leakage through floors and ceilings.
Key points
- A baffle’s percent open area sets resistance to flow, not a fixed quantity of leaked air. The air that actually leaks depends on the pressure difference across the baffle and its surface area.
- Decide where leakage occurs before setting percentages. The presenter points to the rack and containment interface, rack rails, missing blanking panels, and the top and bottom of the rack.
- End caps are normally solid, so the example turns leakage off there.
- In the example, baffle-only containment left roughly 16% more air delivered than the racks needed. Undersized racks with gaps brought that down to about 6%, and the pressure difference between hot and cold sides was clearer.
- More detail means more cells and longer runs. The cabinet-level model ran in 105 minutes against 27 minutes for the simplest version.
- The presenter separates level of detail from accuracy. Detail is what you choose to include, while accuracy comes from the solver.
- Align objects carefully and use the clone tool. Misaligned geometry inflates the mesh and the solve time.
- Make containment large enough that grilles and outlets pass through the baffle with some margin, because the mesher moves objects slightly.