Server room and computer room cooling

Why an IT room is not an office

An office warms up slowly when its rooftop unit quits, and people open a door and complain. A server room holds a steady electrical load around the clock, and nearly every watt that goes into the racks comes back out as heat. With the cooling off, a small closed room climbs past safe inlet temperatures quickly, and the equipment starts throttling or shutting itself down to protect itself.

That load does not drop at night or on weekends, so the cooling cannot either. The same room also needs its humidity held in a band: too wet and moisture condenses where it should not, too dry and static becomes a risk. ASHRAE's thermal guidelines for data processing equipment recommend inlet air of roughly 64–81°F with a moisture ceiling, and most facilities set their alarms well inside that range.

The equipment behind the room

CRAC units (computer room air conditioners) carry their own compressors and reject heat to an outdoor condenser or a fluid cooler, much like a split system built for 24-hour duty. CRAH units (computer room air handlers) have no compressor at all; they run a fan across a chilled water coil fed by a building chiller plant. Smaller rooms often get by on a ductless mini-split head or a dedicated split system, and some telecom closets hang off the building's rooftop unit, which is a problem the first time that unit goes into a night setback schedule.

We service all of it: compressors, condenser coils and fans, evaporator coils, fan motors, humidifiers, condensate pumps and drains, and the controllers that stage it all and raise the alarms. That work on the controllers overlaps with our HVAC controls and control cabinet work, because a large share of computer room calls turn out to be a sensor, a setpoint or a failed changeover rather than a failed compressor.

Computer room faults we see

What you see	What is usually behind it
High temperature alarm, unit running	A dirty or blocked outdoor condenser coil, a failed condenser fan motor, low refrigerant charge, or a load that has outgrown the unit.
High head pressure lockout	Condenser fan or fan motor failure, a coil packed with dirt and debris, or recirculating hot air around a condenser that sits in a corner.
Water under or near the unit	A clogged condensate drain, a failed condensate pump, a frozen evaporator from low airflow, or a humidifier fill valve that stuck open.
Humidity alarm, high or low	A humidifier with a scaled canister or failed element, a reheat stage that is not working, or two units fighting each other over setpoints.
Backup unit never started	A lead/lag changeover that was never tested, a failed transfer relay, or a standby unit that had a problem nobody knew about.

On a computer room call we pull the alarm log first. Those controllers record what happened and in what order, and that sequence usually points straight at the cause.

Redundancy and backup cooling

The industry shorthand is N+1: enough cooling to carry the room (N), plus one more unit so any single unit can fail or be serviced without the room overheating. Plenty of rooms were built that way and drifted out of it. Racks get added, the load climbs, and the "spare" unit quietly becomes a unit the room depends on. A second unit only protects you if it can carry the room by itself and if the changeover actually works.

On a maintenance agreement we test that. We fail the lead unit on purpose, confirm the lag unit picks up the load, check that the temperature and humidity alarms reach the right phone, and log inlet temperatures so there is a trend to compare against next visit. Coils get cleaned, drains get flushed and filters get changed before summer, since the outdoor condensers on these systems work hardest in the same July heat as every other unit in town.

When the cooling fails anyway

The first job on an IT room emergency is keeping the equipment cool, and the repair comes second. The story George Blevens tells most is a TV station's equipment room that was overheating late on a weekend evening after a condenser fan motor failed, with other companies saying nothing could be done until parts arrived. He laid garden soaker hoses across the condenser coil and ran water over it to hold head pressure down, and the station stayed on the air until the new motor went in. That is the kind of thinking we bring to every down room on our 24/7 emergency service.

A better plan is one written before the failure. For a critical room we can help you work out a temporary cooling plan ahead of time: where portable cooling units would sit, where their heat and condensate would go, what power they need, and which door stays propped and which stays shut. Writing it down on a calm Tuesday beats improvising it at 2 a.m.

IT rooms across Middle Tennessee

The region now hosts hyperscale campuses: Google opened a $600 million data center in Clarksville in 2019 (DatacenterDynamics), and Meta's Gallatin campus runs four data centers on about 800 acres (Gallatin Economic Development). Rooms on that scale are a different trade. Most of the region's computing still sits in far smaller rooms, though: the server room in a Clarksville manufacturing plant, the network closet in a Gallatin distribution center, the equipment room behind a studio, a clinic's imaging suite. Those rooms are the ones we work on, in every county on our service area list. The small chillers behind MRI machines are the same kind of job: a compact cooling load that cannot be allowed to overheat, and we service those too.

Computer room questions