Fire suppression interactions with data center HVAC

Fire protection in a data centre has to do two things that can work against each other. It must detect and extinguish a fire quickly, and it must avoid unnecessary shutdowns of IT equipment and the cooling that keeps that equipment alive. HVAC sits in the middle. Air movement can dilute or remove a gaseous agent, open dampers can let it leak away, and an automatic HVAC shutdown can cause the IT room to overheat even if no fire occurs.

Getting these interactions right takes coordination between the fire protection engineer, the mechanical engineer, the electrical engineer and the owner from early design through commissioning. This guide explains the main issues.

Suppression systems used in data centres

The governing standards most often referenced are NFPA 75 (protection of IT equipment), NFPA 76 (telecommunications facilities), NFPA 2001 (clean agent systems) and NFPA 13 (sprinklers). In Canada, the National Building Code and provincial codes set sprinkler requirements for the building, fire alarm systems are installed to CAN/ULC-S524, and the authority having jurisdiction may require specific standards. Owners and insurers also often have their own requirements.

The halocarbon agent landscape is changing. FK-5-1-12 was long sold mainly as 3M Novec 1230, and 3M has exited production as part of its PFAS phase-out, although other manufacturers supply the agent. HFC-227ea is subject to HFC phase-down rules. Owners specifying new systems should discuss long-term agent availability with their fire protection engineer.

Detection and sequencing

Most data centres use very early warning detection, often aspirating smoke detection that samples air continuously through pipe networks. Sampling points are commonly placed at CRAH returns, in hot aisles and above or below raised floors, since airflow carries smoke to these locations quickly. Smoke in a high-airflow room is diluted fast, which is why sensitive detection matters.

A typical clean agent sequence is:

1. Alert level: early detection signals staff to investigate.
2. Pre-discharge alarm: confirmed detection, often from two detectors or zones (cross-zoning), starts a countdown and alarm.
3. HVAC and damper actions: dampers close, outdoor air and exhaust shut off, and HVAC responds according to the agreed sequence.
4. Discharge: agent releases after the countdown, unless aborted.
5. Hold period: the agent concentration must be maintained long enough to extinguish the fire and prevent re-ignition.
6. Post-discharge: purge the agent with ventilation only when the fire service confirms it is safe.

Should the cooling keep running?

This is the most important coordination question. NFPA 2001 generally requires that air handling systems that would reduce agent concentration be shut down or isolated before discharge, unless they are included in the protected volume and do not exchange air with outside.

That leads to a useful distinction:

• Recirculating units inside the protected volume, such as CRAH or CRAC units that draw air from and return it to the same room, do not remove agent. Many designs keep them running because they mix the agent and maintain cooling. Some suppression designers prefer them off during discharge to avoid stratification problems or for other reasons, so this must be agreed explicitly.
• Systems that exchange air with other spaces or outdoors, including outdoor air economizers, makeup air units, exhaust fans and ducted systems serving other areas, must shut down or be isolated by dampers before discharge.

Shutting down all cooling during a pre-discharge alarm can quickly overheat a high-density room. If the alarm is a false one, the result is an IT outage caused by the fire protection system. Thermal effects of any HVAC shutdown should be assessed using the same methods as thermal ride-through analysis.

Air-side economizers

Facilities using outdoor air economizers face a particular challenge. The economizer moves large volumes of outdoor air through the protected space. Before discharge, outdoor air and relief dampers must close and the system must change to full recirculation or mechanical cooling. Damper closing time must fit within the pre-discharge countdown.

Room integrity

Gaseous agents only work if they stay in the room. A room integrity test, usually a door fan test following the method in NFPA 2001, measures leakage area and predicts how long the agent concentration will be held. NFPA 2001 calls for a minimum hold time, commonly 10 minutes unless the AHJ or design requires otherwise.

HVAC components are frequent leakage paths:

• Duct penetrations and unsealed ductwork passing through the enclosure.
• Dampers that do not close tightly or are not fitted where needed.
• Raised floors and ceiling plenums connected to adjacent spaces.
• Cable and pipe penetrations, including condensate and refrigerant lines.
• Pressure relief openings left open by mistake.

Dampers used to isolate the enclosure should be low-leakage and tested. Where they also serve as fire dampers or smoke dampers, they must meet the requirements discussed in fire dampers and smoke control under Canadian codes.

Pressure relief

When a large volume of gas is discharged into a sealed room, pressure changes. Inert gas systems produce a substantial positive pressure; some halocarbon agents can produce a negative pressure early in the discharge as the agent cools the air, followed by a positive pressure. A properly sized pressure relief vent protects the walls, ceiling and doors. Vent sizing is the responsibility of the suppression system designer, but the mechanical engineer must coordinate its location and discharge path and make sure it is not confused with HVAC openings.

Lithium-ion batteries

Lithium-ion battery fires behave differently from ordinary combustible fires. Gaseous agents may suppress flame but cannot stop cell-to-cell thermal runaway, which can generate flammable gases. NFPA 855 and UL 9540A test data guide the design of detection, exhaust and suppression for battery rooms. See battery and UPS room HVAC.

Post-discharge ventilation

After discharge, the space must be cleared of agent and combustion products before re-entry. Many designs include a dedicated purge exhaust system, manually started by authorized personnel, that discharges safely outdoors. Its controls should be clear and separate from normal HVAC.

Integrated testing

Fire suppression, fire alarm, HVAC, dampers, door hold-open devices, electrical power-off functions and the building automation system must be tested as a whole. In Canada, CAN/ULC-S1001 sets out integrated testing of fire protection and life safety systems; see integrated systems testing and CAN/ULC-S1001.

A useful test plan for data centre suppression interactions includes:

1. Verify every HVAC unit, fan and damper response for each alarm stage and zone.
2. Measure damper closing times against the pre-discharge countdown.
3. Confirm that cooling units intended to keep running do so, and that IT temperatures stay within limits during the sequence.
4. Perform the room integrity test with HVAC in its discharge configuration.
5. Confirm the purge sequence and its controls.
6. Document the results so future changes can be checked against them.

Any later change to the room, such as new penetrations, added cooling units or a converted economizer, can invalidate the room integrity test and the sequence. Retest after modifications.

Key takeaways

• HVAC that exchanges air with other spaces or outdoors must shut down or be isolated before a gaseous agent discharges.
• Recirculating cooling units inside the protected volume can often keep running, but this must be agreed explicitly.
• Unnecessary HVAC shutdown on a false alarm can overheat a high-density room.
• Ducts, dampers and plenums are frequent leakage paths that cause room integrity tests to fail.
• Prove the full sequence with integrated testing to CAN/ULC-S1001 and retest after changes.