Skip to content
From Fire Safety Academy Wiki
Mark (talk | contribs)
FSE wiki: Fact-check: qualify observational loss comparisons and 2008 clean-agent exposure conditions (written by Claude Opus 5.5)
Linda (talk | contribs)
Add watermarked Fire Safety Academy illustration at the location specified in the visual plan
Line 21: Line 21:


Most mainframe installations stand on a raised access floor, which carries power and data cables and distributes cooling air.<ref name="Rawson-8-60" /> The floor and its supports should be noncombustible, cable openings should be kept small, and carpet should be avoided.<ref name="Rawson-8-60" /><ref name="Rawson-8-61" /> Raised floors can become accumulation points for trash and debris.<ref name="Mattern-11-31">Mattern (2008b), p. 11-31.</ref> Abandoned cables add fuel in the form of insulation and connectors and block cooling airflow, so a cable management programme should remove them.<ref name="Rawson-8-61" />
Most mainframe installations stand on a raised access floor, which carries power and data cables and distributes cooling air.<ref name="Rawson-8-60" /> The floor and its supports should be noncombustible, cable openings should be kept small, and carpet should be avoided.<ref name="Rawson-8-60" /><ref name="Rawson-8-61" /> Raised floors can become accumulation points for trash and debris.<ref name="Mattern-11-31">Mattern (2008b), p. 11-31.</ref> Abandoned cables add fuel in the form of insulation and connectors and block cooling airflow, so a cable management programme should remove them.<ref name="Rawson-8-61" />
[[File:fse-data-centre-fire-protection-v5.jpg|thumb|upright=1.5|alt=a data centre raised floor with several floor tiles lifted, showing cables and pedestals in the void below|Raised access floor with tiles lifted. The void carries power and data cables and cooling air, and needs its own detection and housekeeping.]]


In New Zealand, Verification Method C/VM2 (second edition, 2025) names floor plenums for IT cables as an example of concealed spaces.<ref name="MBIE-41">Ministry of Business, Innovation and Employment (2025b), p. 41.</ref> Its concealed space scenario applies to buildings with rooms holding more than 50 people.<ref name="MBIE-41" /> The scenario is deemed satisfied by separating elements, automatic detection or suppression.<ref name="MBIE-41" />
In New Zealand, Verification Method C/VM2 (second edition, 2025) names floor plenums for IT cables as an example of concealed spaces.<ref name="MBIE-41">Ministry of Business, Innovation and Employment (2025b), p. 41.</ref> Its concealed space scenario applies to buildings with rooms holding more than 50 people.<ref name="MBIE-41" /> The scenario is deemed satisfied by separating elements, automatic detection or suppression.<ref name="MBIE-41" />

Revision as of 08:52, 8 November 2026

Data centre fire protection is the combination of construction, detection, suppression and emergency planning used to protect rooms and buildings that house information technology (IT) equipment, such as servers, storage systems and network switches. The 20th edition of the NFPA Fire Protection Handbook (2008) treats these spaces as IT equipment rooms, the subject of NFPA 75 in the United States.[1][2] Alongside life safety, business continuity is a central concern: the stored information is often worth more than the hardware, and a short outage can harm a business and its customers.[3][4] Electronic equipment is damaged by heat at temperatures far below those that harm buildings, and also by smoke, corrosive gases and water.[5] Typical protection therefore combines fire-rated separation, very early smoke detection at the ceiling and under the raised floor, sprinklers and often a total flooding gaseous agent.[6][7][8] In US fires in electronic equipment rooms in 1994–1998, the average loss per fire was about three quarters lower where automatic suppression equipment was present than where it was absent.[8]

Hazards and consequences

An IT centre contains more than computers. It includes blade servers, storage devices, tape libraries, routers and switches, together with air-conditioning units, power distribution units and an uninterruptible power supply (UPS).[1] Larger centres add battery rooms, diesel generators, transformers and switchgear in adjacent rooms.[3] The handbook notes that IT centres can also be attractive targets for arson and sabotage, because so much depends on them.[9]

The equipment is very sensitive to heat. Tests summarised in the handbook show that permanent damage can begin above 49 °C (120 °F), and that magnetic tapes and disks start to lose information at 52 °C (125 °F).[5] A sustained ambient temperature of 66 °C (150 °F) damages hard disks, and components begin to fail at 79 °C (174 °F).[5] Major component failures occur between 300 and 500 °F (149 and 260 °C), and plastic tape cases degrade at 650 to 750 °F (343 to 399 °C), releasing flammable styrene monomer.[5][10] These thresholds are far below the temperatures of a developed fire, so equipment can be lost without ever being touched by flame.

Smoke causes a second kind of loss. Tewarson defines nonthermal damage as damage by smoke, toxic and corrosive products, such as corrosion, electrical malfunctions, discoloration and odours.[11] Telephone central offices and computer rooms are among his examples of susceptible occupancies.[12] Burning PVC wire insulation releases hydrogen chloride,[13] and one handbook example sets a design criterion of no more than 5 ppm of it at sensitive equipment.[14] The cable hazard itself is described under Cable fire.

The consequences are mainly economic. In most cases the stored information is more valuable than the IT hardware, so backup copies kept in a safe or remote location are an essential part of protection.[3] The highest level of backup is a fully redundant remote site with mirrored software and real-time data, which is not economically feasible for every owner.[9] Alternatives include contracts with a commercial "hot site", a fully equipped data centre ready to receive the client's software and data.[15] Such planning belongs to business continuity management.

Fire record

In the United States, fires in electronic equipment rooms of non-home structures fell from about 1,400 a year in 1980–1982 to about 900 a year in the late 1990s.[6] From 1999 the national fire data used a narrower code, computer room or control centre, and about 300 fires a year were reported in 2001–2003.[6] Part of this decline may therefore come from the change in coding rather than from fewer fires.[6]

For 1994–1998 the handbook gives an annual average of 944 such fires, with no civilian deaths, 13 injuries and $27.71 million in direct damage.[3] In 133 of these fires no equipment was involved in ignition, in 120 the electronic equipment itself was involved, and fuses, circuit breakers and fixed wiring each accounted for more than 90.[3] Most fires that damage IT rooms start in ordinary equipment in or next to the room rather than in the IT equipment.[3]

About two thirds of the fires, 626 of 944, stayed confined to the object of origin.[8] Even so, the average loss was about $14,600 per fire for those small fires and $29,400 for all fires, which shows how vulnerable the contents are.[5][8] Automatic suppression equipment was present in 379 of the fires; average loss per fire was about three quarters lower than in fires without it.[8] Smoke detection was present in three of five fires and was associated with a two-thirds lower average loss.[7] These comparisons describe the reported fires; they do not isolate the effect of either system from other differences between the buildings.

Location, construction and housekeeping

The handbook recommends that the IT room should ideally be in a separate, single-storey building of fire-resistive or noncombustible construction.[6] Otherwise it may be inside such a building if it is adequately separated, and offices, printers, paper stock and records should be separated from it by rated partitions.[6] For ordinary hazard occupancies the partitions should have at least a 2-hour rating, with listed fire doors and fire dampers in openings and ducts.[6] The room should not be located above, below or next to hazardous processes.[6] Such compartmentation protects high-value operations from a fire elsewhere in the building.[16]

Most mainframe installations stand on a raised access floor, which carries power and data cables and distributes cooling air.[3] The floor and its supports should be noncombustible, cable openings should be kept small, and carpet should be avoided.[3][5] Raised floors can become accumulation points for trash and debris.[17] Abandoned cables add fuel in the form of insulation and connectors and block cooling airflow, so a cable management programme should remove them.[5]

a data centre raised floor with several floor tiles lifted, showing cables and pedestals in the void below
Raised access floor with tiles lifted. The void carries power and data cables and cooling air, and needs its own detection and housekeeping.

In New Zealand, Verification Method C/VM2 (second edition, 2025) names floor plenums for IT cables as an example of concealed spaces.[18] Its concealed space scenario applies to buildings with rooms holding more than 50 people.[18] The scenario is deemed satisfied by separating elements, automatic detection or suppression.[18]

Equipment and services need the same attention. The 2008 handbook says that IT equipment should meet UL 478 or UL 1950, and that automated tape libraries may hold enough plastic to need their own detection and protection.[5] It calls for each computer room to have its own air-conditioning system with noncombustible ducts, able to switch to 100% outside air and exhaust when smoke is detected.[5] It places large battery banks in a separate ventilated room, and standby generators in separate machinery rooms or outdoors with at least one day of fuel.[5]

Detection

The handbook calls for smoke detection at ceiling level and under the raised floor, with underfloor devices listed for that use.[7] Underfloor plenums have high airflow, which makes many detector types unsuitable, so high-sensitivity sampling and spot detectors should be listed for high-airflow conditions.[7] A laser spot detector can be up to 100 times as sensitive as an ordinary light-emitting diode unit and is intended for computer rooms.[19]

Aspirating smoke detection draws air through a pipe network to a central, very sensitive detector.[19] These systems offer several alert and alarm levels and can release a gaseous agent in computer rooms.[19] Their limits are transport time, dilution of smoke over many sampling holes and the fact that airflow at an individual hole usually cannot be sensed.[19] The International Fire Engineering Guidelines (2005) note that such detectors are often up to 10 times as sensitive as point detectors but can respond with a considerable delay.[20]

Air movement is the main difficulty. In the early stages of a fire the plume is weak, so supply and return flows can carry smoke away from detectors or keep it out of them.[21] No quantitative method existed in 2002 for this dilution, but stratification can be estimated.[21] Smoke rises to a height H only if the convective heat release of the fire exceeds a limit set by the temperature difference between the fire location and the ceiling:[22]

q˙conv>0.352H5/2Ts3/2⇔H<(q˙conv0.352Ts3/2)2/5

Symbol Description Unit
q˙conv Convective heat release rate of the fire W
H Height from the top of the fuel to the ceiling m
Ts Temperature difference between fuel location and ceiling °C

The SFPE handbook applies this to overheated PVC cable in a 7 m high air-conditioned room with switch equipment and a temperature difference of 10 °C.[22] For a critical fire of 1,000 W, the smoke rises only about 6 m, so ceiling detectors may not respond.[22] Rearranging the same relation shows that about 1,450 W would be needed to reach 7 m. For a ceiling 3.05 m high, as in a data centre example in the NFPA handbook, about 180 W is enough at the same temperature difference.[23] In low rooms, the dilution of smoke by forced ventilation with high exchange rates is therefore the larger concern.[21]

Sprinklers and water mist

The handbook recommends sprinklers to NFPA 13, hydraulically designed as a wet pipe or preaction system.[8] In a single-interlock preaction system the pipes hold air, and water enters only after a separate detection system opens the valve.[24] In a double-interlock system both the detection system and a sprinkler must operate, and preaction systems are typical in spaces with computer or communications equipment.[25] Rawson advises against the double-interlock type because of its longer delay.[8] The handbook's own performance-based data centre example nevertheless uses a double-interlock system.[23] Few computer rooms meet the definition of light hazard, and rooms with large printers and paper should be treated as Ordinary Hazard Group 2.[8]

The handbook states that there should be little concern about shock or excessive water damage. Experience shows that properly installed and maintained sprinklers cause no measurable increase in damage to electronic equipment compared with heat, smoke and manual hose streams.[26] NFPA 75, as summarised in the handbook, recommends disconnecting power to the equipment before water is applied.[26] Portable extinguishers should be carbon dioxide or Class BC clean agent types, not dry chemical.[5]

A water mist system uses far less water, but its use as a halon replacement in computer rooms has been mixed.[27] In total flooding mode mist cannot extinguish a fire inside an electronic cabinet, whereas a gaseous agent can penetrate the cabinet.[27] One system approved by FM Global draws subfloor smoke into 100 mm tubes, scrubs out soluble acid gases and pushes nitrogen into the cabinets above.[27] Above the floor, mist nozzles at light hazard spacing may match sprinklers on Class A fires with smaller volumes of water.[27] General claims that ceiling mist scrubs smoke are, according to Mawhinney, erroneous.[27]

Gaseous agents

A gaseous extinguishing system is used where the value of the equipment or the risk of interruption justifies it, especially to protect data in process and allow a rapid return to operation.[8] From the 1960s, Halon 1301 total flooding became the usual agent for computer rooms.[28][29] Since the Montreal Protocol it is no longer made in most industrialised countries, and the US Clean Air Act banned new production after 1 January 1994.[30] The 2008 handbook lists nine halocarbon clean agents and four inert gas agents under NFPA 2001 as replacements.[29] It also discusses carbon dioxide systems under NFPA 12, but at about 9% most people become unconscious and at 25 to 30% breathing stops.[8][23] Existing Halon 1301 systems should be tested and maintained to NFPA 12A.[7]

Clean agents are electrically nonconductive, leave no residue and can reach fuel inside closed cabinets.[30][2] If power is cut before discharge, a fire in cable insulation or circuit boards becomes a Class A surface fire, which is relatively easy to extinguish.[2] If power stays on, it remains a Class C fire that is harder to extinguish and can rekindle.[2] The handbook therefore recommends disconnecting power on discharge unless a risk assessment shows less damage with power maintained.[8]

The agent quantity for total flooding follows from the room volume, the design concentration and the specific volume of the agent vapour. For halocarbon agents the mass is given by the first equation below; for inert gases the volume is given by the second, which assumes that leakage during discharge has a rising concentration.[31][32] The halocarbon flooding factor C/(100 − C) assumes that the agent and air lost during discharge are well mixed.[31]

w=VS(C100−C)X=2.303VSlog10(100100−C)VsS=k1+k2T

Symbol Description Unit
w Mass of halocarbon agent required kg
X Volume of inert gas required at 21 °C m³
V Net volume of the protected space m³
C Design concentration % by volume
S Specific volume of the agent at the minimum room temperature m³/kg
Vs Specific volume of the agent at 21 °C m³/kg
k1,k2 Agent constants (HFC-227ea: 0.1268 and 0.0005133; IG-541: 0.659 and 0.00241) m³/kg, m³/(kg·°C)
T Minimum ambient temperature of the protected space °C

As a worked example, take the data centre room of the handbook's performance-based example: 1,858 m² with a ceiling 3.05 m high.[23] Ignoring the floor void and equipment, the net volume is about 5,660 m³. For HFC-227ea at 20 °C, S = 0.1268 + 0.0005133 × 20 = 0.137 m³/kg.[32] At the ISO Class A design concentration of 7.9%, the mass is 5,660 / 0.137 × 7.9 / 92.1 ≈ 3,550 kg.[33]

This is below the agent's 9.0% no observed adverse effect level.[34] The handbook's 2008 table permits a system in a normally occupied space if accidental agent exposure is limited to five minutes; see NOAEL and LOAEL.[35] With IG-541 at its minimum Class A design concentration of 37.5%, about 2,680 m³ of gas at 21 °C is needed instead.[36]

Halocarbon agents must discharge within 10 seconds, mainly to limit the formation of decomposition products.[32] In the handbook's n-heptane liquid-fire tests, the replacement agents formed three to eight times more hydrogen fluoride than Halon 1301, with fire size at discharge the main variable.[32] With automatic detection and actuation, full-scale tests found decomposition levels acceptable in typical computer and electronics spaces.[37] Systems traditionally had to hold their concentration for 10 to 20 minutes.[38]

The enclosure must also survive the discharge. Halocarbon agents first lower and then raise the room pressure, with peaks of about −383 to 383 Pa measured for HFC-227ea.[39] Inert gases and carbon dioxide create a positive pressure that normally requires relief vents.[39] Typical metal stud walls resist over 1,197 Pa, but very tight rooms and rooms over 3.7 m high need care; see Enclosure integrity test.[40] Reliability also matters: for a halon system in a computer facility, the mean probability of failing to protect against a waste-paper fire was estimated at 0.13.[41]

Design approach and emergency planning

Protection works best as layers matched to the stages of a fire. Grant's table for computer rooms links good housekeeping to the pre-ignition stage and smoke detection to initial pyrolysis.[42] Portable extinguishers and an automatic gaseous system cover the incipient stage, sprinklers the pre-flashover stage, and fire walls the post-flashover stage.[42] Each layer protects a different interest, from business operation to the surrounding structure.[42]

The handbook illustrates performance-based design with a corporate data centre whose stakeholders list business continuity first, before life safety and property.[23][4] The example limits interruption of critical operations to 4 hours and requires the agent to extinguish a fire of no more than 2 kW.[43] It also limits halide acids to 200 ppm, requires a hold time of at least 10 minutes and caps discharge pressures at 239 Pa (5 psf).[43] The design fire is sustained arcing in a power circuit that ignites cable insulation inside an equipment cabinet.[44] The example shows a common conflict: owners and insurers may want power to stay on, while code authorities may prefer to cut it before discharge.[45]

An emergency plan completes the protection. It should keep backup copies of software and data, arrange alternative equipment or another site, and invite the local fire service to learn the extinguishing systems.[7] An effective preplan includes a preliminary visit by a restoration company, which can restore much of the damaged hardware.[7] Evacuation procedures should be posted and updated every six months.[7] The handbook also advises protecting telephone switching equipment in the same way as an important computer room; see Telecommunications facility.[46]

Limits and uncertainty

The fire statistics describe US fires reported to municipal fire departments, exclude fires reported only to industrial brigades and use coding that changed in 1999.[6][3] They therefore show trends rather than precise rates. The damage temperatures are general values; components and media differ, and their failure points vary.[5]

The agent equations assume a well-mixed room and a known leakage behaviour.[31] Class A extinguishing concentrations differ between test standards, partly because UL and FM allow 600 seconds for extinguishment and ISO only 180 seconds.[47] Deep-seated fires and cable arrays with long preburn times may need more agent than surface fires.[36] Raising the safety factor does not help if doors are left open.[31]

Detection design also rests on judgement. The SFPE handbook gives no quantitative method for smoke dilution by air handling and notes that smoke reaching a detector does not guarantee that it enters the sensing chamber.[21][22] The requirements cited here are those summarised in the 2008 handbook for US practice, with the NFPA standards of that period. Designers in other countries follow their own codes, such as C/VM2 in New Zealand.[18]

See also

Notes

  1. ↑ 1.0 1.1 Rawson (2008), p. 8-57.
  2. ↑ 2.0 2.1 2.2 2.3 Harrington (2008), p. 17-125.
  3. ↑ 3.0 3.1 3.2 3.3 3.4 3.5 3.6 3.7 3.8 Rawson (2008), p. 8-60.
  4. ↑ 4.0 4.1 Harrington (2008), p. 17-135.
  5. ↑ 5.00 5.01 5.02 5.03 5.04 5.05 5.06 5.07 5.08 5.09 5.10 5.11 Rawson (2008), p. 8-61.
  6. ↑ 6.0 6.1 6.2 6.3 6.4 6.5 6.6 6.7 6.8 Rawson (2008), p. 8-59.
  7. ↑ 7.0 7.1 7.2 7.3 7.4 7.5 7.6 7.7 Rawson (2008), p. 8-63.
  8. ↑ 8.00 8.01 8.02 8.03 8.04 8.05 8.06 8.07 8.08 8.09 8.10 Rawson (2008), p. 8-62.
  9. ↑ 9.0 9.1 Rawson (2008), p. 8-58.
  10. ↑ Rawson (2008), p. 8-61; Celsius values here converted from the printed Fahrenheit values.
  11. ↑ Tewarson (2002), p. 3-133.
  12. ↑ Tewarson (2002), p. 3-136.
  13. ↑ Custer and Milke (2008), p. 2-76.
  14. ↑ Mowrer and Rosenbaum (2008), p. 3-186.
  15. ↑ Schmidt (2008), p. 1-149.
  16. ↑ Davoodi (2008), p. 18-12.
  17. ↑ Mattern (2008b), p. 11-31.
  18. ↑ 18.0 18.1 18.2 18.3 Ministry of Business, Innovation and Employment (2025b), p. 41.
  19. ↑ 19.0 19.1 19.2 19.3 Dungan (2008a), p. 14-22.
  20. ↑ National Research Council of Canada et al. (2005), p. 2.7-6.
  21. ↑ 21.0 21.1 21.2 21.3 Schifiliti et al. (2002), p. 4-26.
  22. ↑ 22.0 22.1 22.2 22.3 Schifiliti et al. (2002), p. 4-27.
  23. ↑ 23.0 23.1 23.2 23.3 23.4 Harrington (2008), p. 17-134.
  24. ↑ Huggins (2008), p. 16-35.
  25. ↑ Huggins (2008), p. 16-37.
  26. ↑ 26.0 26.1 Frank (2008), p. 17-41.
  27. ↑ 27.0 27.1 27.2 27.3 27.4 Mawhinney (2008), p. 16-139.
  28. ↑ DiNenno and Taylor (2008), p. 17-94.
  29. ↑ 29.0 29.1 Harrington (2008), p. 17-123.
  30. ↑ 30.0 30.1 Harrington (2008), p. 17-128.
  31. ↑ 31.0 31.1 31.2 31.3 DiNenno and Taylor (2008), p. 17-113.
  32. ↑ 32.0 32.1 32.2 32.3 DiNenno and Taylor (2008), p. 17-114.
  33. ↑ DiNenno and Taylor (2008), p. 17-98.
  34. ↑ Harrington (2008), p. 17-132.
  35. ↑ Harrington (2008), p. 17-133.
  36. ↑ 36.0 36.1 DiNenno and Taylor (2008), p. 17-112.
  37. ↑ DiNenno and Taylor (2008), p. 17-116.
  38. ↑ DiNenno and Taylor (2008), p. 17-119.
  39. ↑ 39.0 39.1 Harrington (2008), p. 17-126.
  40. ↑ Harrington (2008), p. 17-127.
  41. ↑ Steciak and Zalosh (1992), as cited in National Research Council of Canada et al. (2005), p. 3.7-3.
  42. ↑ 42.0 42.1 42.2 Grant (2002), p. 4-155.
  43. ↑ 43.0 43.1 Harrington (2008), p. 17-136.
  44. ↑ Harrington (2008), p. 17-138.
  45. ↑ Harrington (2008), p. 17-137.
  46. ↑ Transue (2008), p. 9-174.
  47. ↑ DiNenno and Taylor (2008), p. 17-97.

References

  • Custer, R. L. P., & Milke, J. A. (2008). Fundamentals of fire detection. In A. E. Cote, C. C. Grant, J. R. Hall Jr., & R. E. Solomon (Eds.), Fire protection handbook (20th ed., pp. 2-75–2-78). National Fire Protection Association.
  • Davoodi, H. (2008). Confinement of fire in buildings. In A. E. Cote, C. C. Grant, J. R. Hall Jr., & R. E. Solomon (Eds.), Fire protection handbook (20th ed., pp. 18-3–18-21). National Fire Protection Association.
  • DiNenno, P. J., & Taylor, G. M. (2008). Halon and halon replacement agents and systems. In A. E. Cote, C. C. Grant, J. R. Hall Jr., & R. E. Solomon (Eds.), Fire protection handbook (20th ed., pp. 17-93–17-121). National Fire Protection Association.
  • Dungan, K. W. (2008a). Automatic fire detectors. In A. E. Cote, C. C. Grant, J. R. Hall Jr., & R. E. Solomon (Eds.), Fire protection handbook (20th ed., pp. 14-15–14-28). National Fire Protection Association.
  • Frank, J. A. (2008). Characteristics and hazards of water and water additives for fire suppression. In A. E. Cote, C. C. Grant, J. R. Hall Jr., & R. E. Solomon (Eds.), Fire protection handbook (20th ed., pp. 17-31–17-43). National Fire Protection Association.
  • Grant, C. C. (2002). Halon design calculations. In P. J. DiNenno, D. Drysdale, C. L. Beyler, W. D. Walton, R. L. P. Custer, J. R. Hall Jr., & J. M. Watts Jr. (Eds.), SFPE handbook of fire protection engineering (3rd ed., pp. 4-149–4-172). National Fire Protection Association; Society of Fire Protection Engineers.
  • Harrington, J. L. (2008). Application of gaseous agents to special hazards fire protection. In A. E. Cote, C. C. Grant, J. R. Hall Jr., & R. E. Solomon (Eds.), Fire protection handbook (20th ed., pp. 17-123–17-139). National Fire Protection Association.
  • Huggins, R. (2008). Automatic sprinkler systems. In A. E. Cote, C. C. Grant, J. R. Hall Jr., & R. E. Solomon (Eds.), Fire protection handbook (20th ed., pp. 16-29–16-53). National Fire Protection Association.
  • Mattern, L. J. (2008b). Housekeeping practices. In A. E. Cote, C. C. Grant, J. R. Hall Jr., & R. E. Solomon (Eds.), Fire protection handbook (20th ed., pp. 11-27–11-38). National Fire Protection Association.
  • Mawhinney, J. R. (2008). Water mist fire suppression systems. In A. E. Cote, C. C. Grant, J. R. Hall Jr., & R. E. Solomon (Eds.), Fire protection handbook (20th ed., pp. 16-135–16-180). National Fire Protection Association.
  • Ministry of Business, Innovation and Employment. (2025b). Verification Method C/VM2: Framework for fire safety design (2nd ed.).
  • Mowrer, F. W., & Rosenbaum, E. R. (2008). Overview of performance-based fire protection design. In A. E. Cote, C. C. Grant, J. R. Hall Jr., & R. E. Solomon (Eds.), Fire protection handbook (20th ed., pp. 3-183–3-194). National Fire Protection Association.
  • National Research Council of Canada, International Code Council, Department of Building and Housing, & Australian Building Codes Board. (2005). International fire engineering guidelines (2005 ed.). Australian Building Codes Board.
  • Rawson, B. P. (2008). Protection of electronic equipment. In A. E. Cote, C. C. Grant, J. R. Hall Jr., & R. E. Solomon (Eds.), Fire protection handbook (20th ed., pp. 8-57–8-64). National Fire Protection Association.
  • Schifiliti, R. P., Meacham, B. J., & Custer, R. L. P. (2002). Design of detection systems. In P. J. DiNenno, D. Drysdale, C. L. Beyler, W. D. Walton, R. L. P. Custer, J. R. Hall Jr., & J. M. Watts Jr. (Eds.), SFPE handbook of fire protection engineering (3rd ed., pp. 4-1–4-43). National Fire Protection Association; Society of Fire Protection Engineers.
  • Schmidt, D. L. (2008). Emergency management and business continuity. In A. E. Cote, C. C. Grant, J. R. Hall Jr., & R. E. Solomon (Eds.), Fire protection handbook (20th ed., pp. 1-137–1-156). National Fire Protection Association.
  • Tewarson, A. (2002). Generation of heat and chemical compounds in fires. In P. J. DiNenno, D. Drysdale, C. L. Beyler, W. D. Walton, R. L. P. Custer, J. R. Hall Jr., & J. M. Watts Jr. (Eds.), SFPE handbook of fire protection engineering (3rd ed., pp. 3-82–3-161). National Fire Protection Association; Society of Fire Protection Engineers.
  • Transue, R. E. (2008). Fire protection of telecommunications facilities. In A. E. Cote, C. C. Grant, J. R. Hall Jr., & R. E. Solomon (Eds.), Fire protection handbook (20th ed., pp. 9-167–9-177). National Fire Protection Association.