Data center fire suppression systems compared

The comparison, and what each system actually buys you

Data center fire suppression comes down to a short menu of options that protect the same room in very different ways, and choosing among them is a tradeoff between saving the equipment, satisfying the code, and what each system costs you the day it operates. The menu is clean agent gas, pre-action sprinkler, wet pipe sprinkler, and water mist. None of them is the obvious right answer for every room, and the wrong one for a given room can cost more than the fire it was bought to stop.

The fast way to understand the field is to ask one question of each system: what happens when it operates. A clean agent floods the room with gas, puts out a small fire, and leaves the surviving gear dry and clean, but it gets one shot and then the room is unprotected until someone pays to recharge it. A pre-action sprinkler holds water back until a fire is confirmed, then puts water on the fire, which protects the building but wets the gear if it ever flows. Wet pipe is the same water with the pipe charged and waiting over the racks. Water mist splits the difference with far less water.

This guide compares those options head to head and ends with a way to choose. It does not re-teach the door fan test that proves a clean agent room, which the clean agent room integrity guide covers, or the detection and cause-and-effect sequence behind the discharge, which the data center fire and life safety overview covers. The focus here is the decision: which medium over which room, and why.

Why is fire suppression different in a data center?

Fire suppression is different in a data center because the thing you are protecting is destroyed about as easily by the cure as by the fire, and the room runs around the clock with the load live. A data hall is full of high-value electronics that water ruins as thoroughly as flame does, energized continuously, and occupied mostly by equipment with crews in and out rather than a steady population of people. So the suppression has to do three jobs at once: catch a fire early, put it out without wrecking the gear you are saving, and never operate on a false alarm and take the load down by mistake.

That last job is what separates this from an ordinary occupancy. In a warehouse a sprinkler that trips a little early is a wet floor and a cleanup. Over a row of live servers, a single bumped or corroded wet pipe head soaks equipment that was never on fire, and the water loss can run past the loss the fire would have caused. The whole reason the exotic options exist is to avoid that outcome.

The other half is uptime. A clean agent discharge means the cooling shut down, the room is offline, and the recharge clock has started, so even a successful discharge with no fire damage is an expensive, stressful event. The suppression choice is therefore a business-continuity decision as much as a fire one, which is why owners weigh it the way they weigh power and cooling redundancy. The broader fire scheme this sits inside is laid out in the data center fire and life safety overview.

Detection, suppression, and the sprinkler are layers, not a choice

The first mistake people make reading a comparison like this is treating the options as an either-or when the real design stacks them in layers. Early detection is one layer, the clean agent or other primary suppression is a second, and the building sprinkler is a third, and a serious data hall commonly carries all three. They cover different failures. Detection without suppression sees the fire and cannot act on it. Suppression without detection acts late. A clean agent without a sprinkler behind it has nothing left after its one shot.

Think of it as defense in depth tuned to the size of the event. Detection catches the incipient fire, the smoldering power supply or the warm bearing, while it is small enough that a person can pull a circuit and nothing has to discharge. The clean agent handles the small fire that gets past investigation, snuffing it while the gear around it survives. The sprinkler is there for the fire that grows anyway, the developed fire the gas cannot hold, where the job changes from saving equipment to saving the building.

Reading the rest of this guide as a contest between systems misses the point. The comparison that matters is which primary suppression goes over the room, given that early detection sits in front of it and a code-required sprinkler usually sits behind it. The detection layer is the one that decides how often the others ever have to fire.

Why does early detection come first in the comparison?

Early detection comes first because the time it buys is what lets you avoid discharging anything at all, and that changes the value of every suppression option behind it. Aspirating smoke detection, the very-early-warning approach widely known by the VESDA brand, draws air continuously through sampling pipe past a sensitive detector and catches combustion at the off-gassing stage, before there is visible smoke and well before a ceiling spot detector would alarm in a high-airflow room. That early warning is the difference between investigating a problem and reacting to a fire.

The strategy that early detection enables is detect, investigate, then suppress only if needed. A first alarm sends someone to the rack to find the failing component and pull its circuit while the fire is still a smell and a hot smell at that. Most incipient events end there, with no discharge, no downtime, and no recharge bill. The suppression is the backstop for the event that grows past investigation, not the first responder. A scheme that leans on the gas to do work the detection should have done is a scheme that discharges more often than it needs to.

This reframes the whole comparison. With strong aspirating detection in front of it, even a one-shot clean agent rarely fires, because the human in the loop handles the routine event. With weak detection, the suppression carries more of the load and its one-shot limit bites harder. The detection design and sampling strategy are covered in the data center fire and life safety overview; the point here is that it sits ahead of every option below.

What fire suppression do data centers use?

Data centers use one of four suppression mediums over the IT space, usually with one as the gear-saving primary and a sprinkler as the code backstop: clean agent gas, pre-action sprinkler, wet pipe sprinkler, or water mist. The table below is the thirty-second version of the comparison the rest of this guide unpacks.

Read the table as a set of tradeoffs, not a ranking. Clean agent saves the gear but gets one shot and needs a sealed room. Pre-action keeps water out of the pipe until a fire is confirmed, so it tolerates a damaged head without wetting the row, but it still puts water on the gear if it actually flows. Wet pipe is the simplest and the code baseline, and it is the one almost nobody wants charged directly over live racks. Water mist uses a fraction of the water to cool and locally smother, but its design is application-specific and it is less common in IT halls.

The honest summary is that the common data hall pairs a gear-saving primary with a pre-action sprinkler behind it, and the real decision is which primary and whether the sprinkler is pre-action or, where allowed, omitted in favor of the gas. Each row gets its own section below.

What is a clean agent system?

A clean agent system is a total-flooding gaseous fire suppression system that fills a sealed room with an electrically non-conductive agent to a concentration that puts out the fire and leaves no residue behind, governed by NFPA 2001. Nothing has to be wiped down, dried out, or replaced because of the agent itself, which is the whole reason it lives in rooms full of electronics. The gear that was not burning when the system fired comes back online once the room is ventilated.

Clean agents split into two families that put fire out by opposite mechanisms. Halocarbon agents, the chemical agents, work mainly by absorbing heat at the flame faster than the reaction can produce it, and they reach concentration at single-digit percentages while leaving plenty of oxygen for a person to breathe. FK-5-1-12, the low global-warming fluoroketone formerly sold as Novec 1230, and HFC-227ea, sold as FM-200, both work this way. Inert gases, blends of nitrogen, argon, and sometimes carbon dioxide such as IG-541 (Inergen), IG-55, and IG-100, work the opposite way: they lower the room's oxygen from about 21 percent down to roughly 12 to 15 percent, low enough that flame cannot sustain but survivable long enough to leave.

The mechanism drives almost everything else, so it is the first decision and not a detail. The agent families, the design concentration, the human-safety limits, and the cylinder count all follow from whether you picked a heat-absorber or an oxygen-reducer. The room integrity that any of them depends on is covered in the clean agent room integrity guide.

Inert gas vs halocarbon, the environmental clock, and the one shot

Inert gas and halocarbon both extinguish and both leave no residue, so the choice between them turns on storage space, discharge behavior, the recharge model, and the regulatory horizon. Halocarbons reach concentration at single-digit percentages, store compactly, and discharge fast, commonly reaching design concentration in about 10 seconds. Inert gases have to fill a third or more of the room volume, so they carry far more cylinders, often a separate room of them, and discharge more slowly, commonly over roughly 60 to 120 seconds because pushing that volume faster would overpressure the room.

The environmental clock now sits on the halocarbons. HFC-227ea is a hydrofluorocarbon with a high global warming potential, commonly cited around 3,200, and U.S. HFC production is being cut under the AIM Act, stepping down by roughly 85 percent by 2036. An installed FM-200 system is not illegal, but the agent gets scarcer and pricier to recharge, which matters when your protection model assumes you can refill after a discharge. FK-5-1-12 carries a global warming potential near 1, and the inert gases sidestep the phase-down entirely as plain nitrogen and argon with effectively zero global warming potential.

Both families share the limit that defines gaseous suppression: it is one shot. The system discharges once, and from that moment the room has no gas protection until the cylinders are recharged and reconnected. That single fact shapes the cost, the downtime, and the reason a sprinkler usually sits behind the gas. Treat the exact agent concentrations and safety factors as the listing's numbers, verified against NFPA 2001 and the project documents, not as figures to carry in your head.

What is a pre-action sprinkler?

A pre-action sprinkler keeps its piping dry, filled with supervised air instead of water, and admits water only after a separate fire-detection event opens the pre-action valve, so a damaged or corroded head over a rack leaks supervised air and alarms instead of soaking the row. That dry pipe is the reason pre-action, not wet pipe, is the common sprinkler over IT space. The pipe is charged with water only once a fire is confirmed, and only then can a fused head flow.

The interlock logic comes in two forms, and the difference is what it takes to fill the pipe. A single-interlock system admits water when the detection system operates. A double-interlock system admits water only when both the detection operates and a sprinkler head fuses, so neither a false detection alone nor a single failed head alone fills the pipe. Double interlock sounds safer and is sometimes specified for the driest tolerance, but it is slower to deliver water and NFPA 13 penalizes the design with a larger required area of operation, commonly a 30 percent increase for dry-pipe and double-interlock systems. For computer rooms in buildings that otherwise require wet pipe protection, guidance commonly steers toward single-interlock pre-action rather than double for that reason.

Whichever form you use, the interlock is only real if it is trip-tested, and the dry-side designs commonly have to deliver water to the inspector's test connection within about 60 seconds. A pre-action valve that was wired but never trip-tested is an assumption, not a system. The full trip-test method sits with the dry-pipe and pre-action testing covered by topic; the comparison point here is that pre-action is the water option that tolerates a damaged head.

Wet pipe and water mist: the baseline and the fine-droplet option

Wet pipe is the simplest sprinkler there is and the one the building code usually treats as the baseline: the pipe is charged with water at every head, and a head that reaches its temperature rating flows immediately. That immediacy is its virtue everywhere except over live IT gear, where a charged line waiting above the racks is exactly the leak risk the whole data hall design works to avoid. One bumped fitting, one corroded head, one freeze, and water lands on equipment that was never on fire. Wet pipe protects the building reliably and cheaply, and that is why it shows up as the structural baseline even in rooms that also carry gas or pre-action elsewhere.

Water mist sits between the gas and the sprinkler in concept. It discharges very fine water droplets that flash to steam at the fire, cooling the flame and locally displacing oxygen with a small fraction of the water a conventional sprinkler uses, governed by NFPA 750 rather than the sprinkler standard. Less water on the floor means less collateral damage to the gear and the building, which is the appeal. The catch is that water mist design is application-specific and listing-driven, so a system listed for one hazard and room geometry does not transfer freely to another, and it is less common in general IT halls than clean agent or pre-action.

Neither is usually the gear-saving primary in a modern data hall. Wet pipe is the baseline the code leans on, and water mist is the specialty option chosen when its water savings and a supporting listing line up with the room.

Clean agent vs pre-action: which protects the data center?

Clean agent and pre-action protect against different failures, so the real comparison is not which one wins but which one is the gear-saving primary while the other backs it up. Clean agent is the early, asset-saving response: it floods the room, snuffs a small fire while it is still a smoldering supply or a cable fault, and leaves the surviving equipment dry and clean. Pre-action is the building's water protection, the system the code and the insurer rely on to keep the structure from being lost in a fully developed fire that the gas cannot hold.

The genuine tradeoff is this. Clean agent protects the gear and the uptime, but it is one-shot, it needs a sealed room that holds the agent for the required minutes, and once it fires the room is unprotected until it is recharged. Pre-action is reliable, refillable, and code-accepted, and it never runs out the way a cylinder bank does, but it is water on electronics the instant it ever flows. So you are weighing a system that saves the gear on the small event but has a single, expensive shot against a system that will always have water available but means accepting water damage if the protection ever operates.

The lean in the field is to use both and let each do its job: the clean agent for the event you catch early, the pre-action for the event you could not. Removing the sprinkler because there is a gas system is how a contained incident becomes a lost building, and removing the gas because there is a sprinkler trades the gear away on every small fire. Where budget forces one, the code and the AHJ usually settle it, and the answer is most often the sprinkler, because life safety and structural protection outrank equipment protection.

How detection and suppression work together on discharge

Whichever primary you pick, it is wired to the detection system through a releasing sequence built so a single faulty detector never operates the suppression, and the sequence is the same shape across the gaseous and pre-action options. Detection is cross-zoned: two independent detection zones have to agree before the system arms. A first detector raises the alarm and sends someone to investigate. A second, independent detector confirms it, and only then does the releasing sequence start. That two-detector logic is what keeps a dusty smoke head from discharging a six-figure cylinder bank or filling a pre-action pipe.

Once both zones confirm, a pre-discharge time delay runs before a gaseous agent releases, commonly on the order of 30 seconds, with a horn and strobe warning the room. The delay gives people time to leave and gives an operator who knows it is a false alarm time to hit the abort station, which holds the discharge while the cause is dealt with. There is also a manual release for someone who sees the fire and will not wait for the detectors. The emergency power off, the EPO that drops the room's electrical load, is a separate control with its own button, and the design has to settle how it interacts with the discharge so that killing power and releasing agent do not fight each other.

All of this lives in the cause-and-effect matrix under NFPA 72, and it is only real if the integrated test proves it: trip both zones, watch the delay run, test the abort and the manual release, confirm fans stop and dampers close, and confirm the panel actually energizes the release at the end of the count. The full detection and cause-and-effect treatment is in the data center fire and life safety overview.

Does a data center still need sprinklers with clean agent?

In most jurisdictions, yes: the building code usually still requires a sprinkler system even when a clean agent protects the room, and the clean agent is treated as added protection rather than a replacement for the sprinkler. Many authorities having jurisdiction will not accept a gaseous system as the only automatic suppression in a data center, and the common arrangement is a pre-action sprinkler installed in addition to the clean agent. NFPA 75, the standard written for IT equipment rooms, points toward sprinkler protection for these spaces, and the clean agent is the layer on top of it.

The reasoning follows from the one-shot limit. A gas system protects against a fire it can reach and hold while the room is sealed and the cylinders are charged. It does not protect against a fully developed, ventilation-fed fire, and after it discharges it has nothing left until it is recharged. The sprinkler is the protection that is always available and that handles the large event, so the code keeps it whether or not a gas system is present. The gas saves the asset on the small fire; the sprinkler keeps the building standing on the large one.

Do not promise an owner that the clean agent lets them skip the sprinkler. That is a conversation for the AHJ and the adopted building code, not the brochure, and the adopted edition and local amendments control the answer. Confirm what the jurisdiction requires before the design assumes the gas stands alone, because designing around an omitted sprinkler that the AHJ then requires is an expensive correction late in the job.

People safety: agent discharge, oxygen, and egress

A suppression comparison has to account for the people who might be in the room when it operates, because the gaseous options change the air the way water does not. Halocarbon clean agents are designed so the concentration stays at or below the no-observed-adverse-effect level for a normally occupied room, so the agent is safe for the brief expected exposure while the fire is out and the air is still breathable. Inert gases work by lowering oxygen to roughly 12 to 15 percent, low enough to stop flame but survivable long enough to leave promptly, which is why the egress design assumes people clear the room on the pre-discharge alarm.

The pre-discharge sequence is the safety system around the discharge. A confirmed alarm starts a time delay, commonly on the order of 30 seconds, with a horn and strobe, so the room can be cleared before any agent releases, and the abort station lets someone hold the discharge while they deal with a known false alarm. The discharge itself is loud and disorienting, especially for inert gas, so signage, the warning devices, and a clear path out are part of the design, not an afterthought. Nobody should be trapped in a room that is about to flood with gas.

Water-based systems carry a different and smaller human risk, mostly the slip and the electrical hazard of water near energized gear, which is one more reason the room is de-energized appropriately as part of the sequence. The agent that saves the equipment is not something to be standing in, so confirm the exposure limits against the agent listing and confirm egress against NFPA 101 and the AHJ. The full egress treatment is in the data center fire and life safety overview.

The enclosure: why the gaseous options need a sealed room

The deciding weakness of every gaseous option is that it only works if the room holds the agent, and that makes the enclosure part of the system rather than a container for it. A clean agent reaches extinguishing concentration in seconds, then starts leaking back out through every gap in the room, and the moment the concentration falls below what the fire needs to stay dead, the fire can reflash. Water stays where it lands. Gas does not. So a gaseous system carries a requirement that the water options do not: the room must be sealed and proven to hold the concentration for a minimum time, commonly at least 10 minutes under NFPA 2001.

Holding the room means the fans shut down and the dampers close on the fire signal, because a running air handler sweeps the agent out and pulls fresh air in faster than any cylinder bank can replace it. It means cable and conduit penetrations, door gaps, and the raised floor and ceiling plenum are sealed, because the protected volume includes the space under the floor and often above the ceiling. And it means a pressure relief vent, because sealing the room tight enough to hold the agent also means the fast discharge has nowhere to dump its pressure spike, so a too-tight room can be damaged by its own discharge.

All of that is proven by the room integrity door fan test before the system is accepted and repeated periodically after, because rooms do not stay sealed once people start pulling cable through them. That test, the sealing, and the pressure venting are the subject of the clean agent room integrity guide. The comparison point is blunt: the water options do not care whether the room is sealed, and the gaseous options fail completely if it is not.

Maintenance, the one-shot problem, and recharge

The cost comparison does not end at installation, because the systems differ sharply in what they cost to keep ready and what they cost after they operate. A clean agent system has the one-shot problem: after a discharge, the room is unprotected until the cylinders are recharged and reconnected, and the recharge is expensive, especially for the phase-down halocarbons whose agent is getting scarcer. A discharge with no fire damage at all still means cooling shut down, the room offline, the cylinders empty, and a bill to refill them, which is why an accidental discharge is one of the events operators fear most.

The cylinders also carry ongoing checks that the water systems do not. Halocarbon cylinders are weighed on a cycle, often every 6 months, because the agent is a liquid and weight tells you how much is left; inert gas cylinders are judged by pressure off the gauge. A slow valve leak empties a cylinder over months with no alarm, and a bank that discharges half empty never reaches design concentration. On top of that the cylinders are pressure vessels with a hydrostatic retest clock under DOT rules, independent of the fire code, and a cylinder past its retest date cannot legally be refilled, which becomes a problem the day you need to recharge.

A pre-action or wet pipe system, by contrast, is refillable from the building water supply and does not run out after one operation. It carries its own maintenance under NFPA 25, the valve trip tests, the main drain test, the fire pump flow test, but it does not have a one-shot limit or a recharge bill in the same sense. When you compare lifecycle cost, the gas wins on saved equipment and loses on recharge and the one-shot exposure; the water wins on always-available and refillable and loses on the water damage when it flows.

How do you choose a data center fire suppression system?

Choose the data center suppression system by working through the things that actually constrain the decision, in roughly this order: what the code and the insurer require, how much equipment value and uptime are at stake, how much downtime the operation can tolerate, the room size and the cylinder space available, and the budget for both install and recharge. The code requirement usually sets the floor, a sprinkler, and the rest of the decision is which gear-saving primary, if any, sits on top of it.

Run it as a set of questions rather than a preference. Is a sprinkler required by the adopted code and the AHJ? Almost always yes, so a pre-action sprinkler is the likely baseline. Is the equipment value and the cost of downtime high enough to justify a clean agent on top, with its install and recharge cost and its sealed-room requirement? For most serious data halls, yes. If clean agent, halocarbon or inert, which turns on cylinder space, discharge noise near spinning media, the recharge model, and the regulatory horizon on the agent. Is there a case where water mist's lower water volume and an application-specific listing fit better than either? Sometimes, but it is the exception, not the default.

The decision matrix below lines the options up against the factors that decide them. Use it to frame the conversation with the owner and the AHJ, not to replace either, because the adopted code, the insurer's data sheets, and the project's basis of design control the final call.

High-density AI racks and liquid cooling

High-density racks and the move to liquid cooling are changing the fire picture faster than the comparison above was written for, so treat this area as evolving and lean on the manufacturer, the AHJ, and the fire protection engineer rather than habit. AI and high-performance compute have pushed per-rack power far above the levels older detection and suppression layouts assumed, which concentrates more energy and more potential fuel in a smaller footprint and changes how a fire would start and spread. The aspirating detection sampling strategy and the suppression coverage that suited a 5 kW rack are not automatically right for a rack drawing many times that.

Liquid cooling adds a new variable. Direct-to-chip and immersion cooling bring coolant, manifolds, and in the immersion case dielectric fluid in open or sealed tanks, into the same space as the suppression, and the interaction between a coolant leak, the cooling fluid's own fire behavior, and a gaseous or water-based discharge is not something to assume from the air-cooled case. A dielectric immersion fluid behaves very differently in a fire than the air it replaced, and the suppression has to be considered against the actual cooling architecture, not a generic data hall.

There is no settled, one-size answer here yet, and the standards are catching up to the hardware. The safe posture is to design the fire protection around the specific cooling architecture and rack density, confirm the approach with the AHJ and the insurer, and document the basis of design clearly so the next reviewer knows what assumptions the suppression was sized against. The cooling architecture itself is covered by topic in the data center cooling material.

The battery and energy-storage room is a different problem

A lithium-ion battery or energy-storage room is a different fire problem than a data hall, and the suppression that protects the IT space does not automatically protect it. Lithium-ion failures involve thermal runaway: a failing cell vents flammable gas, makes its own heat, and can propagate cell to cell, producing a fire that reignites and that a clean agent flood will knock down but not necessarily stop, because the chemistry keeps generating heat and fuel from inside the pack. A gaseous agent sized to put out an electronics fire is not a reliable answer to a runaway battery.

NFPA 855, the standard for stationary energy storage systems, is the governing document, and it pulls in detection, suppression, gas detection for the flammable off-gas, explosion control, spacing, and a hazard mitigation analysis. The deflagration risk from vented battery gas is a real explosion hazard that needs ventilation and explosion control, which is a different set of provisions than a clean agent room carries. The standard applies above threshold quantities that vary by chemistry and occupancy, so a small UPS battery may fall below it while a real battery energy storage system does not.

The practical lesson for the comparison is to scope the battery room separately and not assume the data hall's clean agent or pre-action covers it. Confirm the threshold, the adopted edition, and the required protections with the AHJ, because NFPA 855 is moving quickly. The battery room protection is covered by topic in the energy-storage material and in the data center fire and life safety overview; the point here is that it does not ride on the IT room's suppression choice.

What to document

The suppression comparison has to end in a documented decision, because the question after any incident, or any later expansion, is what was installed, why, and what role it plays in the code scheme. Capture the medium chosen and the agent or water type, whether it leaves residue, whether it is one-shot or refillable, and what role it plays under the code, the gear-saving primary or the required baseline. The table below is the minimum record of the decision.

Common mistakes

• Treating a clean agent system as a replacement for the sprinkler, leaving the structure unprotected on a developed fire.
• Specifying a clean agent and then ignoring room integrity, so the agent leaks out a cable trench before it can extinguish.
• Choosing a suppression medium with no early detection in front of it, so the system discharges on events a person could have caught.
• Designing without an abort station, a pre-discharge time delay, and a cleared egress path, so people are exposed to the discharge.
• Putting a wet pipe sprinkler charged with water directly over live racks where a single damaged head soaks the row.
• Picking a one-shot gaseous system without planning and budgeting the recharge and the downtime after a discharge.
• Assuming the data hall's clean agent or pre-action also protects the lithium-ion battery room, which needs NFPA 855 protections.
• Promising the owner the gas lets them skip the sprinkler without confirming it with the AHJ and the adopted building code.
• Carrying agent concentrations or code section numbers from memory instead of the listing and the adopted edition.
• Sealing a clean agent room tight for retention and providing no pressure relief vent, so the discharge damages the enclosure.

Field checklist

Standards and references

The suppression options each answer to their own standard, and the occupancy standards sit over all of them. NFPA 75, the Standard for the Fire Protection of Information Technology Equipment, covers the data center room and points toward the protection it expects, and NFPA 76 covers telecommunications facilities. Clean agent systems are governed by NFPA 2001, the Standard on Clean Agent Fire Extinguishing Systems, which sets the agent concentration, the retention the room integrity test confirms, and the pressure relief and discharge criteria. The agent's own listing fixes the agent-specific numbers, the extinguishing concentration, and the human-exposure limits.

Water-based systems, including pre-action and wet pipe, are installed to NFPA 13, the Standard for the Installation of Sprinkler Systems, while water mist falls under NFPA 750 and is not treated as a sprinkler system under NFPA 13. Detection and the releasing sequence behind any of these follow NFPA 72, the National Fire Alarm and Signaling Code. After turnover, the water-based systems are inspected, tested, and maintained to NFPA 25, and the clean agent cylinders carry their own checks plus the DOT hydrostatic retest clock under 49 CFR. Stationary battery energy storage is governed by NFPA 855.

Edition numbers and clause references change every cycle, the jurisdiction adopts a specific edition that is often a few cycles back, and it amends locally, so confirm the adopted editions and any local amendments against the project documents before citing a clause, and let the AHJ settle any conflict. The insurer's FM Global data sheets often impose criteria stricter than code and govern by contract where the carrier requires them. The room integrity proof is detailed in the clean agent room integrity guide, and the full fire scheme in the data center fire and life safety overview.

Units, terms, and acronyms

Fire suppression comparison carries vocabulary that reads differently across an NFPA 2001 calculation, a sprinkler submittal, and an insurer data sheet. The terms below are the ones that travel across the whole decision.

Clean agent

A gaseous fire-extinguishing agent that leaves no residue, either a heat-absorbing halocarbon or an oxygen-lowering inert gas

Total flooding

Filling an entire sealed enclosure with agent to a design concentration, rather than aiming at a single object

Halocarbon vs inert gas

Halocarbons (FK-5-1-12, HFC-227ea) put fire out mainly by absorbing heat; inert gases (IG-541, IG-55, IG-100) lower oxygen below combustion

Pre-action

A sprinkler system whose pipe stays dry until a detection event opens a valve, so a damaged head does not leak water

Single vs double interlock

Single admits water on detection alone; double requires both detection and a fused sprinkler head before water enters the pipe

Wet pipe

A sprinkler system charged with water at the heads, ready to flow on a fused head; the code baseline, avoided over live racks

Water mist

A system of very fine water droplets that cool and locally smother with far less water than a sprinkler, governed by NFPA 750

One-shot / recharge

A gaseous system discharges once and is then unprotected until the cylinders are recharged and reconnected

GWP

Global warming potential; high for HFC-227ea (around 3,200), near 1 for FK-5-1-12, effectively zero for inert gases

AHJ

Authority having jurisdiction, the official who adopts the editions, approves the design, and settles conflicts

Related tools

Calculators and readiness checks for this work

Compare your options

FAQ

People also ask

• What is a door fan test?
• What fire suppression do data centers use?
• What is a battery monitoring system?
• What is an integrated systems test?
• Cast-in vs post-installed anchors: which should I use?