Skip to content

Insights

What to monitor in a server room, and where each sensor goes

8 min read

What to monitor in a server room comes down to six things: the temperature of the air entering the equipment, humidity, water on the floor, the mains supply, the alarm outputs of the UPS and plant already in the room, and the door. We supply all six as one complete system: sensors in ZARC tags, a ZARC Gateway and the dashboard that sends the alerts. Each one needs its sensor in a particular place, and most of the value is lost when a sensor goes where it is easy to fit rather than where the fault shows first. This guide covers each measurement, where it goes, which alerts are worth setting and how to get the readings out of a room that is already full and running.

01

Inlet temperature: the reading the equipment cares about

IT equipment is specified against the temperature of the air at its inlet, not the temperature of the room. ASHRAE's TC 9.9 thermal guidelines, which equipment makers design to, recommend an inlet range of 18°C to 27°C for most servers and network equipment. A wall thermostat measures room air, usually by the door, and can read 22°C while the top of a rack draws in air several degrees warmer.

Fit the sensors on the front of the racks. Warm exhaust air that recirculates over the top of a rack or around its sides shows first at the top of the front face, so one sensor at the top of each rack front catches most problems. On racks that run hot, add a second at mid-height. Missing blanking panels and gaps between racks let exhaust air back to the front, and a top-of-rack reading that sits well above the others is often the first sign of it.

Set two limits and a rate. A warning a few degrees below 27°C gives time to act, and an alarm at 27°C marks the edge of the recommended range. A rate-of-rise alert matters more than either in a small room: when an air conditioning unit stops, inlet temperatures climb quickly, and a rise of several degrees in a few minutes is a cooling failure long before any fixed limit is crossed.

02

Humidity and dew point

Humidity matters less than temperature in most small rooms, but it is cheap to measure alongside it. The same ASHRAE guidance sets the recommended moisture range by dew point, from -9°C to 15°C, with an upper relative humidity limit of 60% in the fourth edition and 70% in the fifth where the air is clean. Above that, condensation becomes a risk on cold surfaces such as chilled water pipework and the coils of the cooling units. Very dry air raises the risk of static discharge.

A combined temperature and humidity sensor at inlet height on one rack per room, or one per aisle in a larger room, is enough. Read humidity as a trend: a steady climb usually points to a cooling unit that has stopped removing moisture, or to water getting in from outside.

03

Water: monitor where leaks start, not the middle of the floor

Leaks in server rooms come from a short list of sources: the cooling units and their condensate drains and pumps, pipework running through or over the room, and the roof. A leak sensor works by detecting water where it touches the sensor, so it has to sit where water will arrive first. Put one on the floor directly under each cooling unit and condensate pump, one under each run of pipework, and, in a room with a raised floor, one under the floor tiles near each of those, since water can spread across the slab for some time before it is visible.

A small sealed water leak tag sits on the floor with no cable to run, which makes it simple to place under plant and inside floor voids. A sensor in an open, dry part of the floor adds little.

04

Mains power and the UPS

A UPS hides a power cut. When the mains fails, the UPS takes the load without a break and the room carries on, so nothing looks wrong until the batteries run down and the equipment shuts off. The useful alert is the loss of mains itself, raised the moment it happens, while the UPS still has battery time left for someone to act.

Monitor the supply upstream of the UPS, on the circuit that feeds the room or the UPS input, with a power failure tag that runs on its own battery and keeps reporting when the mains has gone. Restoration matters too, so the alert history records when the supply came back and how long the outage lasted. Most UPS units also have volt-free alarm contacts for on battery, low battery and fault. Connecting those adds the UPS's own view of its state to the same record.

05

Doors, and the equipment that already raises alarms

A magnetic door contact on the inside of the room door gives a timestamped record of every opening. It shows who was in the room around the time of a fault, and an alert on an out-of-hours opening is a simple access check for rooms with no access control of their own.

Most rooms also contain equipment that already knows when it has a problem: the air conditioning controller, the fire suppression panel, a standby generator. These usually have volt-free alarm outputs that nobody is watching. A dry contact tag reads those outputs without modifying the equipment, so a cooling fault or a suppression discharge raises an alert in the same place as a temperature alarm. Where an old auto-dialler currently carries those alarms, the same approach is how a GSM alarm dialler is replaced.

06

What to monitor in a small server room: a starting layout

The table below is a starting point for a single room. Larger rooms add inlet sensors rack by rack and leak sensors source by source; the principles stay the same.

What to measureWhere the sensor goesAlert worth setting
Inlet temperatureFront of each rack, top position first; add a mid-height sensor on the hottest racksWarning a few degrees under 27°C, alarm at 27°C, rate-of-rise for a cooling failure
HumidityAt inlet height, on one rack front per room or per aisleHigh humidity, and low humidity in rooms with very dry winter air
WaterOn the floor under cooling units, condensate drains and pipework, and under a raised floorAny water detected, at once
Mains powerOn the supply feeding the room or the UPS, upstream of the UPSLoss and restoration, at once
UPS and plant outputsOn the volt-free alarm contacts of the UPS, air conditioning and suppression panelOn battery, low battery, fault, suppression discharged
DoorMagnetic contact on the inside of the room doorOpened out of hours, or left open

07

Getting the readings out of the room

Retrofitting cabled sensors into a live room means cable runs, containment and often a shutdown window. Battery-powered wireless sensors avoid all three, but the radio has to get out of a room full of metal racks, frequently in a basement or the middle of a building. Lower frequencies carry through walls and around obstructions better than higher ones, which is why sensors in this kind of space work better at 433 MHz than at 2.4 GHz. Test from the real sensor positions before committing to any system.

The connection back is the second check. Monitoring that relies on the network it is watching goes quiet when that network fails, which is often when it is needed most. A gateway with its own 4G connection and battery backup keeps alerts flowing through a network fault or a power cut, and keeps the monitoring devices off the corporate network.

The third is security. Sensors that only transmit, and take no instructions, give nothing to attack on the radio side and cannot be used to control anything in the room. That is the reasoning behind one-way telemetry, and it makes the security review for a new monitoring system much shorter.

08

Alerts, escalation and silence

Every reading should go to a dashboard, and the dashboard should raise the alert and send it by SMS or email to named people, with escalation to the next person if nobody acknowledges it. Decide who receives which alert before go-live: a leak under a cooling unit and a door opened at 2am usually need different people.

Silence needs an alert of its own. Each sensor should check in on a schedule, every six to ten minutes as standard on our tags, so a sensor that stops reporting is flagged rather than mistaken for a quiet room. The gateway should do the same, so a loss of the gateway is never invisible.

For a room or an estate of comms rooms, the complete picture of these measurements is set out on our server room monitoring page.

FAQ

Frequently asked questions

What should be monitored in a server room?

Six things cover most server and comms rooms: the temperature of the air entering the equipment, humidity, water on the floor under cooling units and pipework, the mains supply feeding the room or UPS, the alarm outputs of the UPS and other plant, and the door. Each reading goes to a dashboard, which sends alerts by SMS or email to named people.

Where should a temperature sensor go in a server room?

On the front of the racks, where the equipment draws in air, starting at the top of each rack front because recirculated hot air shows there first. A sensor on the wall or by the door measures room air, which can be several degrees cooler than the air some equipment is taking in.

What temperature should a server room be?

ASHRAE's TC 9.9 guidelines recommend an inlet air temperature of 18°C to 27°C for most IT equipment, measured at the front of the racks. Set a warning a few degrees below 27°C and an alarm at it, and add a rate-of-rise alert to catch a cooling failure early.

Where should leak sensors go in a server room?

On the floor directly under the sources of water: each cooling unit, condensate drain and pump, and each run of pipework, plus under a raised floor near those sources. A leak sensor only detects water that reaches it, so placement matters more than the number of sensors.

Why monitor the mains if the room has a UPS?

Because a UPS takes over silently. Without monitoring upstream of the UPS, nobody knows the mains has failed until the batteries run down and the equipment shuts off. An alert on the loss of mains gives the time the UPS provides to someone who can use it.

Can server room monitoring be fitted without cabling?

Yes. Battery-powered wireless sensors fix in place with no data cabling or mains supply, report to a gateway on site and can be added one room at a time. Test the radio from the real sensor positions first, since racks and building structure affect range.

Make your site signals visible, usable and evidenced.

Tell us about the sites, assets and conditions you need to monitor. We will help scope a practical pilot for reliable telemetry, real-time alerts and evidence-ready reporting.

Location

Venator House, Unit 9, 15-17 St Stephen's Road, Bournemouth

Dorset, BH2 6LA · VAT GB 409644484

Tell us about your enquiry

Provide a few details about your sites, assets and operational needs. We will be in touch to discuss the right approach.

We'll only phone you if you choose phone as your preferred contact.

Preferred contact (optional)

Optional. Leave blank and we reply by email.

No mailing lists. No spam. We reply directly to your enquiry.