A data centre is not an office, and it is certainly not a warehouse. The combination of high air velocities, hot/cold-aisle architecture, dense cable trays, and zero tolerance for downtime means that conventional point detectors in a server hall are almost always too late.
In this article: why standard smoke detection falls short in data- centre environments, which detection strategies NPR 5310 prescribes, and how to engineer the BMI/OAI without undermining business continuity.
1. Why point detectors fail in cold aisles
A conventional optical smoke detector reacts to smoke particles that diffuse, in sufficient concentration, to the sensor element. In a hot/cold-aisle configuration, however, air is actively moved — often at several metres per second. Smoke particles disperse extremely quickly and reach the detector at very low concentration, so the alarm only triggers once the fire is already substantial.
In practice a point detector that would react within 30 seconds in a normal office can take several minutes to react in a cold aisle. For a data centre that is too late — damage and downtime are already inevitable.
2. The norm: aspirating detection per NEN-EN 54-20
For mission-critical environments practice prescribes aspirating smoke detection (ASD) — known under brand names such as VESDA, Stratos and FAAST. These systems actively draw air through a pipe network above the zone and analyse it in a central detector block. Because the system samples actively, sensitivity is up to 100× higher than a passive point detector — a very early fire indication, well before the human nose registers anything.
NEN-EN 54-20 defines three classes:
- Class A — very high sensitivity, required for mission-critical server rooms
- Class B — elevated sensitivity, suitable for technical rooms with mechanical ventilation
- Class C — comparable to conventional point detection
For a server hall with cold-aisle containment, Class A is the norm.
3. NPR 5310 as a practice guide
For data-centre engineering, NPR 5310 offers a concrete reference: detector layout, pipe routing, sampling holes, alarm levels and integration with BMS, EPO and cooling. ANiX engineers data-centre BMI/OAI per this NPR, combined with manufacturer-specific design tools (VESDA Pipe Design, Stratos PipeIQ).
Points from NPR 5310 that often matter most in practice:
- Above and below the floor. Hot/cold aisle plus sub-floor cable management requires detection on both levels
- In the return-air path. ASD sampling in the return-air ductwork catches smoke before it is diluted in the hot aisle
- Hot-rack detection. Per-rack detection on critical racks reduces search time when an alarm goes off
- Pre-alarm staging. Class A allows multiple alarm levels (Aware, Action, Fire1, Fire2) — coupled to a graduated response instead of immediate evacuation
4. Integration without downtime
A data centre is not taken offline for BMI/OAI installation. ANiX works inside data centres to a fixed protocol:
- Project intake with site reliability / facility teams — what are the change windows, which racks are no-touch, what procedures govern access to the white space
- Pre-fabrication of pipe networks outside the white space — the in-room installation requires as little dwell time as possible
- Integration with the existing BMS — alarm output to SCADA / building management uses the same protocols (BACnet, Modbus, Contact-ID) already in use for cooling and power
- Functional tests during an announced change window — smoke generators per manufacturer specification, witnessed by client and inspection body
Closing
Fire safety in a data centre is an engineering discipline, not an installation task. The standard is clear (NPR 5310, NEN-EN 54-20 Class A) but execution requires experience with air flow patterns, BMS integration, and colocation SLAs.
Working on a new or existing data centre where the BMI is dated? Schedule a sector intake — we look at your layout, cooling strategy and current installation.