Data center fire protection: ensuring business continuity

Figure 1. Cost of Data Center Outages [Ponemon Institute, 2016].

The occurrence of a fire in a data center can lead to business interruption (“downtime”) costs in the hundreds of thousands of US dollars. According to the most recent Ponemon Institute report, the average cost of a data center outage in 2016 was $740,357; on a time basis, the average cost of data center downtime was $8,851 USD per minute.

The two major costs of downtime are those associated with business disruption – the total economic loss of the outage, including damage to reputation and lost business opportunities – and lost revenues – the total revenue loss from customers and potential customers because of their inability to access core systems during the outage. The total cost of downtime is not limited to these losses, but will ultimately include the cost of

User productivity loss: lost time and expenses associated with end-user downtime
IT productivity loss: lost time and expenses associated with IT personnel downtime
Detection cost: associated with the discovery and investigation of the outage
Recovery cost: costs related to restoring networks and systems
Ex-post response cost: after-the-fact costs due to the business interruption
Equipment cost: cost of new equipment purchases and repairs
Third-party cost: cost of contractors, consultants, auditors and other specialists

Figure 1 summarizes the cost of a typical data center outage and shows that these additional costs can be significant.

Fire Protection Options

The high value and sensitivity of the electronic equipment found in data centers, combined with the consequences of system interruption, make fire protection a crucial component of any data center risk assessment. Fires do occur in these facilities as evidenced in Table 2, which lists a selection of data center fires reported within the last several years.

A. Structure Protection

Most data centers are required by code to have a sprinkler system installed for protection of the structure against fire. The primary objective of a sprinkler system is not fire extinguishment but fire control: confining a fire to its point of origin (preventing fire spread) and controlling ceiling temperatures to prevent structural damage.

Water has obvious disadvantages around electronics and electrical systems due to its electrical conductivity. In the event of sprinkler system activation, water damage to the facility and equipment can be substantial, often worse than the fire damage itself, and the cleanup required can be extensive. In addition, water is not a three-dimensional agent, and cannot readily extinguish hidden or obstructed fires, such as an in-cabinet or in-rack fires. For these reasons, sprinkler systems are best suited for the protection of structures, not for the protection of mission-critical assets located within those structures.

B. The Minimalist Approach to Data Center Fire Risk Mitigation

A common minimalist approach to fire protection in data centers is to install sprinklers for the protection of the structure and high sensitivity smoke detectors (HSSDs) for asset protection. The theory behind this approach is that once a fire is detected, someone can find the fire and extinguish it, perhaps with a handheld extinguisher. This approach requires 7×24 manning of the facility, and failure on the part of the operator to find and extinguish the fire could obviously lead to disastrous results impacting both the operator and the facility.

The potential consequences of adopting a minimalist approach to data center fire protection can be clearly seen in the devastating results of a recent fire at the Shaw data center in Calgary, Canada. The fire protection system in this case consisted of the minimum protection required by code, i.e., a sprinkler system. The impact of this fire included:

Knockout out of the primary and backup systems supporting key public services
Loss of cable, telephone and Internet services by more than 20,000 Shaw business and household clients
Crippled city services, including 311 emergency services
Delay of hundreds of surgeries at local hospitals
ATMs and debit terminals throughout the city affected
Extensive water damage to equipment on the floors below the top story fire location
Temporary relocation of over 900 Shaw employees while damage is repaired
Six days of service outage

According to media reports, an electrical fire triggered the facility’s sprinkler system which ran for more than two hours, soaking furniture, walls and sensitive electronic equipment on floors below. The total cost of the incident was not limited to the costs associated with the above items, but ultimately also included costs due to loss of data and records, lawsuits and the loss of customer confidence.

C. Asset Protection

The protection of sensitive and expensive electronics requires the use of a clean fire extinguishing agent. The primary objective of a gaseous clean agent system is to extinguish the fire quickly, limiting fire damage to the object(s) involved in the origin of the fire. Hence, the primary purpose of a gaseous clean agent system is to protect the valuable, sensitive and mission-critical assets within the enclosure. This is clearly fundamentally different from the primary objective of sprinkler systems, as illustrated in Figure 2.

Figure 2. Fire Control vs Fire Extinguishment.

Clean agent systems employ a combination of rapid detection and rapid agent discharge, providing extinguishment of fires in their incipient stage. This significantly reduces asset damage due to thermal effects or fire combustion products, allowing facilities to quickly return to service after a fire event. Furthermore, clean agents do not leave corrosive or abrasive residues following their use, eliminating the cost and need for cleanup as well as the potential for longer term equipment operational issues. A gaseous clean agent penetrates hidden or obscured areas and densely packed cabinets and racks. Consequently, clean agent systems are ideally suited as the first line of defense to protect electronic equipment in data centers.

The four key value elements to be considered in the selection of a clean agent are:

Speed of Extinguishment
Space and Weight Reduction
Safety Margin
Sustainability/Environmental Warranties

Table 2 provides a summary comparison of the four key value elements for the three most widely employed clean agents FM-200, Novec 1230, and Inergen.

Table 2. Comparison of Clean Agents: Class C Fires.

Speed of Extinguishment

For Class C hazards, FM-200 and Novec 1230 systems are required to reach 95% of their minimum design concentration (MDC) within 10 seconds. Inergen systems must reach 95% of the required MDC within 120 seconds – as a result, extinguishment times with Inergen systems can be much longer than those for FM-200 or Novec 1230 systems.

Space and Weight Requirements

Inert gas agents cannot be compressed to liquids and can only be stored as high-pressure gases. As a result, inert gas extinguishing systems require the use of large numbers of high pressure storage cylinders, and hence require much more storage space compared to compressed liquefied agents such as FM-200 or Novec 1230.

As seen in Table 3, compared to FM-200 systems, Novec 1230 systems require 20% more mass for Class C hazards of less than 480 volts, and 57% more mass for Class C hazards involving voltages equal to or exceeding 480 volts. As a result, the weight and space requirements for Novec 1230 systems exceed those for FM-200 systems.

Table 3. Clean Agent Requirements: Data Center. a NFPA 2001 (2018); b FM Data Sheet 4-9 (2019) Protection.

Safety Margin

Inert gas agents such as Inergen do not undergo chemical reactions within the body and as a result are of very low toxicity. The No Effect Level (NEL) and Lowest Effect Level (LEL) for Inergen are 43% and 52%, respectively [NFPA 2001, 2018 edition].

Like Inergen, FM-200 does not react in the human body and hence is of very low toxicity. Novec 1230 is chemically reactive, and when inhaled reacts with water in the lungs. As seen in Table 4, FM-200 offers a higher safety margin/safety in use compared to Novec 1230, as it has lower toxicity vs Novec 1230, as measured by a number of toxicological effects, including short term inhalation (4h LC50), cardiac sensitization lowest observed adverse effect level (CS LOAEL), physiologically-based pharmacokinetics (PBPK), chronic inhalation (28-day inhalation), allowed exposure levels (AELs) and metabolism.

Sustainability

Inergen does not contribute to ozone depletion or global warming and hence is environmentally benign. FM-200 does not contribute to ozone depletion and has a negligible impact on global warming: based on US EPA data, the impact on global warming of all HFCs in fire protection applications represents less than 0.02% of the impact on global warming of all greenhouse gases [US EPA, Inventory of US GHG Emissions & Sinks, 2018]. Novec 1230 does not contribute to ozone depletion, and like FM-200 has a negligible impact on global warming.

The manufacturers of both FM-200 and Novec offer 20-year environmental warranties as an assurance of the sustainability of the products. Chemours offers their 20-year Falcon and 20-year Eagle Programs for FM-200, and 3M offers their 20-year Blue Sky Warranty for Novec 1230. These warranties provide a guarantee that the agents are long-term solutions.

Table 4. Comparison of FM-200 and Novec 1230 Toxicology. Yellow indicates lower toxicity by the indicated toxicological measure. a NFPA 2001 (2018); b DuPont FM-200 SDS; c NICNAS Report STD 1019, p. 10; d Chemours FM-200 PUSH Bulletin; e 3M Novec 1230 SDS; f Armstrong Laboratory Report AL/OE-TR-1995-0022 (1995); g Novec 1230 Fire Protection Fluid Safety Assessment [3M]

Ensuring business continuity

To ensure business continuity in data centers, protection of both the structure and its contents is required. The added cost of installing a clean agent system is justified by its ability to provide what sprinkler systems cannot – protection of the sensitive, expensive and mission-critical assets located within the facility, and the minimization or complete elimination of business interruptions in the event of a fire.

For more information, go to FM200.com