Server rooms rarely fail loudly. The air conditioner trips, the room warms by a degree every few minutes, and the first sign is a slowdown or a reboot long after the cause. This page describes a temperature and humidity monitoring setup for server rooms, comms rooms and small data centres: sensors that watch the room, alerts that reach IT on call, and a record that helps explain what happened.
Why server rooms overheat quietly
Most in-house server rooms started as a store room with an extra air conditioner. The cooling is sized for the equipment on day one, and then more racks, a UPS and a switch stack arrive. The air-conditioning unit is a single point of failure, and the room has no thermostat other than the one on that same unit. When it stops, a small room full of equipment can pass the safe limit surprisingly fast. In Jeddah the effect is made worse by the building, because the warm surrounding spaces offer no help when the cooling goes down.
There are slower failures too. A unit can keep running while its capacity drops because of a dirty filter or low refrigerant, and the room settles at a temperature that is merely too high. Dust, condensation and high humidity cause corrosion and short circuits, while very dry air raises the risk of static discharge. None of these is visible to staff who walk past the door.
Temperature, humidity, water and door sensors
A server room needs more than a temperature number. HHET builds the set around the failures that occur in practice:
| Sensor | What it detects | Typical placement |
|---|---|---|
| Temperature | Cooling failure, hot spots | Rack front at mid-height and top, plus return air |
| Humidity | Too damp or too dry | One per room or aisle, away from vents |
| Water leak | Pipe, drain or condensate leak | Under raised floor, near AC unit and pipework |
| Door contact | Unplanned access, door left open | Room entrance and rack doors where used |
| Power status | Mains loss before the UPS runs down | On the monitoring gateway or a power input |
Placement is the part most often done badly. A sensor on the wall near the door reads the corridor, not the equipment. Equipment draws cool air in at the front of the rack, so the useful reading is at the intake, at the top of the rack where hot air rises, and at the exhaust side to see how much heat the equipment is adding. Two or three sensors per rack row give a far more honest picture than one in the middle of the room.
A temperature that is acceptable at the floor can be well above the limit at the top of a rack. If you only put in one sensor, put it near the top of the rack intake.
Alerts to IT on-call
An alarm that sounds in an empty room is only a noise. The HHET platform sends alerts to the people who can act: app notification and SMS to the IT on-call, escalation to a second person if nobody acknowledges, and a different alert level for a warning versus a critical alarm. At a warning, someone can check the AC; at a critical level, someone can begin an orderly shutdown of non-essential systems before hardware is damaged. The rules for who is called and when are set with you, and they can be changed in minutes when the rota changes.
Rate-of-change alerts are worth enabling. If the room gains several degrees in ten minutes, it is a cooling failure even if the absolute temperature is still within range. That early warning is often worth the whole system. For a wider look at alarm types, escalation and sensor choice, see the page on temperature and humidity sensors.
Monitoring needs its own resilience. A sensor network that stops when the building loses power is no use during a power cut, so gateways are fitted with battery backup and, where the corporate network is part of the problem, a cellular uplink. The monitoring system should not live on the same network switch that sits in the room it watches.
What a good response looks like
Monitoring is only half the answer; the other half is the procedure behind the alert. For a cooling failure, write down the order of actions: check the unit and its power, open the door and use a portable fan if the room can tolerate it, shut down lower-priority systems first, and call the maintenance contractor. Put the contractor's number in the alert message itself. After the event, use the graph to measure how long the room took to reach each threshold; that tells you how much time you really have next time, and whether to move the warning limit lower. Over a year, the history also shows seasonal patterns, such as the room struggling on August afternoons, which supports a case for added cooling before the failure happens.
Recommended ranges
The equipment manufacturer's specification comes first. Industry guidance, such as ASHRAE's published recommendations for data centres, commonly gives an intake air range in the region of 18 to 27 °C, and a humidity range expressed as dew point and relative humidity, but you should check the current figures and your hardware's own limits. Many small rooms are simply kept at around 22 to 24 °C for comfort and margin.
| Measure | Common practice | Notes |
|---|---|---|
| Rack intake temperature | About 18 to 27 °C | Follow equipment specification; narrower for older hardware |
| Relative humidity | Moderate, avoiding both very dry and condensing conditions | Check current guidance and manufacturer limits |
| Rate of rise | Alert on several degrees in minutes | Detects cooling failure early |
| Water | Any detection | Zero tolerance, immediate alert |
Set the warning a few degrees below the point where the equipment would throttle or shut down, because you need time to act. The guide to server room temperature and humidity standards explains the published ranges in more detail.
Because the platform keeps the history, it also answers the question that follows every incident: when did the room start to warm up, and did the alert reach the right person? That record is useful for fault tickets, insurance and reviewing whether cooling needs upgrading. Sensors can be added later as racks are added. To specify a system for your room, request a quote with a floor plan or a photograph and the number of racks.
If the same business also runs cold storage or a warehouse, those sensors can sit on the same platform. Server rooms are the simplest case of environmental monitoring, but the principles are the same ones used for cold rooms: measure where it matters, alert the right person, keep the record.

Last reviewed .
