Size from the measured IT electrical load, not from room dimensions. Use equipment rated for continuous duty that restarts after a power interruption. Provide redundancy with duty rotation, and alarms that reach a person remotely.
Sizing from the IT load
Essentially all the electrical power drawn by IT equipment is converted to heat in the room. That is the load, continuously, regardless of the weather.
- Measure the actual electrical drawAt the distribution board or the rack PDUs. Do not use equipment badge ratings – they are nameplate maximums and are typically far above real draw.
- Add the UPS lossesA UPS is not 100 per cent efficient and its losses are heat in the room.
- Add lighting and occupancySmall, and not zero.
- Add fabric and solar gainIf the room has an external wall or a window. A comms room with a south-facing window is a design error worth correcting.
- Add a growth allowanceA room sized exactly to today’s load has no headroom, and adding capacity later is disruptive.
- Decide the acceptable room temperatureCheck the equipment manufacturers’ stated operating ranges rather than defaulting to a remembered figure – modern equipment tolerates considerably warmer conditions than traditional practice, and over-cooling is a large continuous cost.
A great many comms rooms run far colder than the equipment requires, because somebody set a number years ago. Raising the setpoint to within the manufacturers’ stated ranges reduces running cost continuously and reduces the load on the cooling plant, which improves its reliability too.
Why a comfort unit is risky
Standard comfort air conditioning is common in small comms rooms because it is cheap and available. It carries specific risks that are worth understanding before choosing it.
| Issue | Why it matters in a comms room |
|---|---|
| Designed for intermittent duty | A comms room load is constant, twenty-four hours a day, all year |
| Defrost or anti-freeze cycles | Many comfort units periodically stop cooling to defrost. In a room with a constant load, that is a temperature rise every cycle |
| Does not restart after power interruption | Many default to off. After a power cut, the UPS keeps the servers running and the cooling does not restart. The room then heats with nothing to stop it |
| Control strategy built for human comfort | Not for holding a tight band against a constant load |
| Low-ambient operation | A comms room needs cooling in winter too, and not all comfort units run well at low outdoor temperatures |
| No alarm output | No way to tell anybody it has failed |
Where budget dictates a comfort unit – and frequently it does – specify one that restarts automatically after power interruption, that will operate at low outdoor ambient, and add independent monitoring. Those three additions remove most of the risk.
Redundancy that works
The question is not whether the cooling will fail, but what happens when it does.
| Arrangement | Gives you | Suits |
|---|---|---|
| Single unit | Nothing. Failure means the room heats up | Only where the room can genuinely be shut down |
| N+1, standby on failure | A second unit takes over | Better than single, and the standby is unproven |
| N+1 with duty rotation | Both units share runtime and each is proven regularly | The right answer for most business comms rooms |
| Separate electrical supplies | Survives a circuit failure as well as a unit failure | Where the room genuinely matters |
| 2N | Full duplication | Data centre territory |
Duty rotation deserves emphasis. A standby unit that sits idle for two years will very likely not start when it is finally needed – seals dry out, the condensate path blocks, and nothing has proven otherwise. Rotating duty weekly or monthly means both units are continuously proven and both wear evenly. It costs nothing to configure and it is the difference between redundancy and the appearance of redundancy.
Monitoring, which is the cheapest resilience
- Temperature sensors at the rack inlets, not on a wall by the door. The inlet temperature is what the equipment actually experiences, and it can differ substantially from ambient.
- High temperature alarm, set to give time to respond rather than at the point of damage.
- Unit failure alarm from the cooling equipment itself.
- Water detection under the units and along the condensate path. A condensate leak into a rack is a different kind of disaster from overheating.
- Remote notification that reaches an actual person – text, email, or through an existing monitoring platform. An alarm that only sounds in an empty building is decoration.
- Escalation, so an unacknowledged alarm reaches somebody else.
- Test it. An alarm system nobody has tested since installation is an assumption rather than a control.
Where the organisation already has IT monitoring, tying the cooling alarms into it is usually easier and more reliable than a separate system, because that is the platform people already watch.
Airflow and the room itself
- Cold air must reach the rack inlets and hot exhaust must return to the cooling unit. In a small room this mostly means not blocking the front of the racks and not letting hot exhaust recirculate into the inlets.
- Rack orientation. Racks facing each other front-to-front creates a cold aisle; back-to-back creates a hot aisle. Racks all facing the same way means every rack breathes the one in front’s exhaust.
- Blanking panels in empty rack U-spaces. Without them, hot exhaust air bypasses straight back to the inlet. They cost very little and measurably reduce inlet temperatures.
- Seal the room. Cooling a room with a propped-open door cools the corridor.
- Cable management. A dense mat of cables at the rear of a rack obstructs exhaust airflow.
- Condensate. Gravity drainage, routed away from anything electrical. Where a pump is unavoidable it needs its own failure alarm.
- Access for maintenance, which in a room full of racks is frequently an afterthought.
Comms room on a single comfort unit with no alarms? That is the most common arrangement and the least resilient one.
Questions we get asked about this
Can I use a normal air conditioning unit in a server room?
It is common and it carries risks – comfort units are not designed for continuous full-load duty, many have defrost cycles that interrupt cooling, and many default to off after a power interruption. If budget dictates one, specify auto-restart, low-ambient operation, and add independent monitoring.
How cold should a server room be?
Less cold than most people assume. Check your equipment manufacturers’ stated operating ranges – modern equipment tolerates considerably warmer inlet temperatures, and over-cooling is a large continuous cost.
Do I need two units?
If the room failing would genuinely disrupt the business, yes – and with duty rotation rather than a permanent standby, because a standby that has never run is not a standby.
What happens in a power cut?
The UPS keeps the servers running and the cooling usually stops, so the room heats rapidly with the full IT load still producing heat. It is worth knowing how long you have, and worth confirming the cooling restarts automatically when power returns.
How will we know if it fails?
Only if something tells you. Temperature and failure alarms with remote notification to an actual person are the single cheapest piece of resilience in the room.
What are blanking panels for?
They fill empty rack spaces so hot exhaust air cannot bypass straight back to the equipment inlets. They cost very little and measurably reduce inlet temperatures.
Need this doing?
Call and describe the problem — you will speak to an engineer, not a call centre. Written quotes, no estimates over the phone.