Server Room Monitoring — Temperature, Humidity, Smoke, Leaks

In short

ASHRAE recommends 18–27 °C for IT equipment; allowable ranges depend on equipment class and span 15–32 °C (class A1) to 5–45 °C (class A4).
Humidity has two limits: too low raises electrostatic discharge risk, too high causes corrosion. ASHRAE contamination guidelines call for keeping relative humidity below roughly 60%.
Rate of change matters too: IT equipment tolerates at most 20 °C per hour and no more than 5 °C in any 15-minute window.
Uptime Institute: cooling causes 19% of outages, power 52%, and human error contributes to between two thirds and four fifths of all downtime incidents.
Temperature alone is not enough — a complete picture needs humidity, leak detection, early smoke detection and cabinet door monitoring.

What temperature a server room should hold

The reference point is the ASHRAE thermal guidelines for data processing environments. They distinguish a recommended range for everyday operation from allowable ranges that depend on equipment class.

The recommended range is shared across classes A1–A4 and runs 18–27 °C. Allowable ranges are as follows:

Allowable ranges per ASHRAE TC 9.9 (2021 edition)
Class	Temperature	Max dew point
A1	15–32 °C	17 °C
A2	10–35 °C	21 °C
A3	5–40 °C	24 °C
A4	5–45 °C	24 °C
H1 (high-density servers)	15–25 °C, recommended 18–22 °C	17 °C

Class H1, introduced in the fifth edition, deserves attention. For high power density servers the recommended range is distinctly narrower at 18–22 °C. Rising rack power density therefore means less headroom, not more.

Rate of change can matter more than the value

ASHRAE limits dynamics as well as values: IT equipment tolerates at most 20 °C per hour, and no more than 5 °C in any 15-minute window. For tape media the limit is stricter — 5 °C per hour and humidity change below 5% per hour.

This translates directly into monitoring design: a system that reads temperature once an hour will not catch a rapid rise after a cooling failure. A measurement interval in the order of minutes is not excessive — it is the condition for noticing an event that actually threatens the equipment.

Humidity: two limits, two different risks

Humidity in a server room is a two-sided problem, and therefore harder than temperature.

Too dry: electrostatics

Low humidity raises the risk of electrostatic discharge. Citing research by Pommerenke and others, ASHRAE reports that in data centres without ESD flooring, where staff may wear non-ESD footwear, the probability of generating an 8 kV discharge rises from 0.27% at 25% relative humidity to 0.43% at 8% — roughly a 1.6-fold increase. Under such conditions the guidelines require a grounded wrist strap for any work on IT equipment.

Too damp: corrosion and leakage currents

ASHRAE contamination guidelines state that relative humidity in a data centre should be kept below roughly 60%. Above that, settled dust starts absorbing moisture, which drives corrosion and leakage currents — and those currents rise exponentially with relative humidity.

The document also defines corrosion thresholds corresponding to a modified G1 severity level per ISA-71.04: copper corrosion below 300 angstroms per month and silver below 200 angstroms per month. If coupon testing stays within those limits, operation up to 70% humidity is acceptable; if it exceeds them, humidity should be lowered to 50% or less.

All ASHRAE ranges are defined as non-condensing. In practice what matters is the margin between surface temperature and dew point, not the relative humidity figure on its own.

What to monitor first

The order follows from how quickly each phenomenon damages equipment and how hard it is to notice without a sensor.

Air temperature at rack inlets. That is what determines operating conditions — not the temperature in the middle of the room or near the ceiling.
Water leaks. Water destroys immediately and irreversibly, and it spreads across the floor where nobody is looking.
Relative humidity. Slower to cause harm, but expensive in consequence: corrosion and electrostatics.
Smoke. Early detection of smouldering buys time to react before flames appear.
Cabinet and door opening. Access control and evidence in case of an incident.

It is worth remembering that Uptime Institute attributes between two thirds and four fifths of all downtime incidents directly or indirectly to human error, and four in five respondents believed their last major outage was avoidable with better procedures. Monitoring does not replace procedures, but it gives them a foundation: without measurement there is nothing to follow and nothing to review.

Setting alarm thresholds that do not flood the inbox

The most common mistake is one threshold for everything. An alarm that fires several times a day stops being read within a week — and the one that mattered is lost with it.

Separate warning from alarm. Warn on approaching the edge of the recommended range; alarm on leaving the allowable range for the equipment class.
Add a time condition. A brief excursion when the door opens is not a failure. One that persists for fifteen minutes is.
Set a separate threshold on rate of change. A jump of several degrees in a quarter of an hour signals a cooling failure even while the absolute value is still within range.
Test alarms periodically. A notification channel nobody has verified in a year is a channel that does not work.

A formal framework for data centre infrastructure is provided by the EN 50600 series, matching international ISO/IEC 22237, whose environmental control part classifies designs by availability and protection class.

Leaks and smoke: the two costliest events

NFPA 75, the standard for fire protection of information technology equipment, offers a recommendation worth knowing at design stage: steam, water and horizontal drain piping other than sprinkler piping should not be located above suspended ceilings or above IT equipment. The standard also advises against basement locations for server rooms, and where equipment does sit in a basement, requires measures against flooding from internal and external sources — including firefighting water from floors above.

That yields a list of places worth covering with leak sensors: beneath the raised floor, next to precision cooling units, near water and drainage risers, and by the room entrance if the server room lies below ground level.

With smoke, airflow is the problem. NFPA 75 notes that detection systems find flaming fires more readily than smouldering ones, and that the greater the airflow, the poorer detection performance, because combustion products are diluted and channelled away. At very high air change rates, detectors under a raised floor may fail to catch a fire originating beneath that floor without abnormally dense spacing — which is why the standard recommends locating detectors in return air streams and at exhaust grilles.

What the absence of monitoring costs

The most credible figures come from the Uptime Institute annual outage analysis. The breakdown of causes of the most recent significant outage looks as follows:

Causes of significant outages per Uptime Institute (2024)
Cause	Share
Power	52%
Cooling	19%
Third-party provider	9%
IT systems (hardware and software)	8%
Network	7%
Fire and suppression	3%
Information security	1%

On cost: 54% of respondents reported that their most recent significant outage cost more than 100,000 US dollars, and 16% more than a million. At the same time 55% of operators experienced an outage in the past three years, though only one in ten was serious or severe.

These figures cover downtime cost rather than equipment replacement, so the real cost of an event is usually higher. It is also worth noting that water leaks are not a separate category in the Uptime data — which does not mean they do not happen; they sit inside the cooling and power categories, because that is where their consequences show up.

Frequently asked questions

What temperature should a server room be?

ASHRAE recommends 18–27 °C for IT equipment. Allowable ranges depend on equipment class: from 15–32 °C for class A1 to 5–45 °C for class A4. For high power density servers (class H1) the recommended range is narrower, at 18–22 °C.

What humidity should a server room have?

ASHRAE contamination guidelines call for keeping relative humidity below roughly 60%. Operation up to 70% is acceptable if corrosion coupon testing shows copper below 300 and silver below 200 angstroms per month. Humidity that is too low, in turn, raises electrostatic discharge risk.

What should be monitored first in a server room?

Air temperature at rack inlets, then water leaks, relative humidity, smoke, and cabinet or door opening. The order follows from how quickly each phenomenon damages equipment and how hard it is to notice without a sensor.

How often should server room temperature be measured?

ASHRAE allows IT equipment a change of up to 20 °C per hour but no more than 5 °C in any 15-minute window. To catch an event within those bounds, the measurement interval should be in the order of minutes. Hourly readings will not support a response to a cooling failure.

Does server room monitoring require the cloud?

No. A logger can serve data locally through its built-in web page, SNMP and JSON or XML readout, and send alerts by e-mail directly. The cloud is useful for remote access and viewing several sites at once, but it is not a precondition for monitoring to work.

Sources

Need monitoring at your site?

We will help you choose sensors and the way they run — locally on your LAN, in your own monitoring system or in the CE MonitorApp cloud.

Contact us See products

Server room monitoring: temperature, humidity, smoke and water leaks