Hot Aisle Containment in a Mixed-Density Hall
A 600 kW hall has 12 rows of cabinets. Eight rows run 4-6 kW per rack; four rows were recently populated with 15 kW GPU nodes. Return-air temperatures at the CRAHs climb to 32°C, and three GPU racks trip thermal alarms weekly. Perforated tile airflow was set years ago for uniform 5 kW racks. There is no containment; cold and hot aisles mix openly at row ends. The CRAH units have spare capacity, and chilled-water supply temperature is 12°C. The facilities team proposes adding two more CRAHs. Budget is tight and the floor is nearly full.
1. Is adding CRAH capacity the right first move? Justify your recommendation with the airflow and mixing evidence.
Read the worked answer
Probably not first. The symptom is return air at 32°C with spare CRAH capacity available, which suggests the problem is distribution and mixing, not total cooling capacity. Uniform perforated tile airflow tuned for 5 kW racks delivers roughly the same cold air to a 15 kW rack as to a 5 kW rack, so the high-density rows are starved while low-density rows are over-supplied. Open row ends allow hot exhaust to recirculate into cold aisles, raising inlet temperatures on the very racks that need the coldest air. Adding CRAHs without fixing airflow would add fan energy and may worsen mixing. The first moves are containment (hot or cold aisle) and rebalancing tile airflow toward the high-density rows, then re-measure rack inlet temperatures. Adding capacity should be a later decision if, after containment and balancing, CRAH return temperatures and rack inlets still show insufficient total airflow.
2. What measurements would you take to confirm the diagnosis and verify the fix?
Read the worked answer
Measure rack inlet temperatures at top, middle, and bottom of each high-density rack (not just CRAH return), and compare to the 15 kW racks' airflow demand. Measure CRAH supply and return temperatures and airflow to compute actual delivered cooling versus capacity. Use a hot-aisle/cold-aisle differential: a large gap between cold-aisle and hot-aisle temperatures indicates good separation; a small gap indicates mixing. Check perforated tile open area and airflow at tiles in front of GPU racks versus low-density rows. After containment and rebalancing, re-measure the same points; success looks like rack inlets within the equipment's allowable range and CRAH return temperatures rising toward design, with no thermal alarms.