Your Cage Runs Hot at 60 Percent Load. It Is Almost Never the Cooling.
Recirculation, missing blanking panels, and blocked rear exhaust cause more thermal incidents in colocation cages than facility cooling capacity ever does. Here is how to find and fix them.
A recurring call: the cage is nowhere near its contracted power draw, but inlet temperatures on the top third of two cabinets are climbing, and something has already thermally throttled. The instinct is to escalate to the facility about cooling capacity. In our experience the facility is almost never the cause.
The cause is airflow management inside the cabinet — specifically, cold air reaching equipment inlets without first mixing with hot exhaust. That is a problem you own, and it is usually fixable in an afternoon.
The one principle everything else follows from
Data center cooling is not about making a room cold. It is about keeping two air masses separate: the cold supply going into equipment fronts, and the hot exhaust leaving equipment rears. Every thermal problem inside a well-built facility is a leak between those two masses.
When hot exhaust finds a path back to the cold aisle, it mixes with supply air, raises the inlet temperature, and the equipment's own fans respond by pulling harder — drawing in yet more of the warm air. It is a feedback loop, and it concentrates at the top of cabinets because hot air rises.
The five leak paths, in the order we find them
1. Empty rack units with no blanking panels. An open U is a direct tunnel from the hot rear to the cold front. This is the single most common finding and the cheapest to fix. Blank every unused U, including the odd 1U gap left "for the next server."
2. Unsealed cable cutouts. The brush grommet or floor cutout under the cabinet, left open, dumps supply air where nothing needs it, or lets exhaust travel underneath. Seal cutouts with brush strips or pillows.
3. Gaps between cabinet and rails. The vertical channels beside the mounting rails are a path from rear to front unless side blanking or air-dam kits are fitted. On wide cabinets with dense cabling, this leak is significant and almost never addressed.
4. Cable congestion at the rear. A wall of unmanaged cabling behind the equipment restricts exhaust, raising back pressure and internal temperatures even when inlet air is fine. Rear cable managers and horizontal routing solve it; zip-tied bundles across the exhaust face do not.
5. Open cabinet sides or missing doors. In a shared row, an open side panel lets your cabinet exhaust into a neighbour's cold aisle, or theirs into yours. Perforated front and rear doors are correct; solid ones or removed ones are not.
Measure delta-T, not room temperature
Room temperature is nearly useless as a diagnostic. The two numbers that matter are the inlet temperature at each cabinet — measured at the front face, at three heights: bottom, middle, and top — and the difference between inlet and exhaust, the delta-T.
What the readings tell you:
- Top inlet significantly warmer than bottom inlet: recirculation over the top of the cabinet, or through open U space. Look for missing blanking panels first.
- Delta-T unusually low across the cabinet: too much air is bypassing the equipment entirely. You are cooling the room, not the servers, and paying for it.
- Delta-T unusually high: insufficient supply volume reaching that cabinet, or restricted exhaust. Check floor tile placement and rear cable congestion.
- All inlets rising together across the row: this one may genuinely be the facility. Now escalate, with data.
Walking in with a temperature map at three heights per cabinet changes the conversation with a facility from an argument into a diagnosis.
ASHRAE's recommended envelope, and what it actually permits
Industry thermal guidance for typical enterprise equipment allows a considerably wider inlet range than most operators assume — modern equipment tolerates inlet temperatures well above the traditional cold-room habit. The point is not that you should run hot. The point is that raising supply temperature is safe and efficient only when airflow is properly separated. In a cage with recirculation, raising supply temperature makes the top of the cabinet worse immediately.
Fix separation first. Efficiency gains come after, not before.
Containment: what you can and cannot do as a tenant
Full hot-aisle or cold-aisle containment is usually a facility-level design. As a tenant in a cage or a partial row, you generally cannot rebuild containment — but you can do the tenant-side equivalents:
- Blank every open U, every time, as part of the install checklist rather than as a later clean-up.
- Fit side air dams or brush kits in the rail channels.
- Keep front and rear doors closed and perforated.
- Match equipment orientation so everything in a cabinet pulls front-to-rear. Any device installed backwards — a switch with reversed airflow is the classic case — turns one cabinet into a mixing chamber. Order airflow-direction-specific switch models to match the cabinet.
- Seal floor cutouts and remove tiles that were opened for a deployment that finished years ago.
That last item is worth a walk-through of its own. Perforated tiles migrate. Someone opens one during an install and never closes it, and supply pressure quietly drops for the rest of the row.
Density changes the problem, not the principle
Higher-density deployments — dense compute, GPU nodes, populated storage shelves — concentrate heat in a way that magnifies every leak. A cabinet drawing modest power tolerates sloppy airflow. The same cabinet at high density does not: the same missing blanking panel that cost a couple of degrees now costs many.
Before deploying dense hardware into existing space, confirm three things with the facility, in this order: available power per cabinet, available cooling for that heat load in that specific row, and whether the containment design of the row supports it. Power is frequently available before cooling is, and the deployment that fits electrically can still cook itself.
A one-afternoon remediation
For a typical cage that has grown organically over a couple of years:
- Photograph the front and rear of every cabinet before touching anything.
- Record inlet temperature at three heights per cabinet.
- Blank every open U.
- Seal floor cutouts and remove stray open tiles.
- Dress rear cabling off the exhaust face into vertical managers.
- Fit side air dams where rail channels are open.
- Close and latch all doors.
- Re-measure the same points after a settling period, and record the delta.
The before-and-after temperature map is the deliverable. It is also the evidence you need if the remaining gap really is a facility issue.
What good looks like
Every U accounted for — occupied or blanked. No visible path from rear to front. Cabling dressed vertically and clear of the exhaust. Doors closed. A current elevation drawing that matches reality. Inlet temperature within a few degrees from bottom to top of every cabinet.
StackTrue performs airflow and thermal audits as part of rack and stack work and as a standalone engagement, and hands back a before-and-after temperature map with the elevation documentation. If a cage is throttling below its contracted capacity, that is usually the fastest cost saving available on the floor.
Planning a colocation, enterprise, or AI infrastructure project?
Get a free, vendor-neutral consultation from our Canadian data center team. No contract, no pressure — just an honest operational read on your options.
