An Ounce of Prevention Keeps Data Center Fires at Bay
August 01, 2007

By Staff
Appeared in Building Operating Management/FacilitiesNet

In 1965, Intel co-founder Gordon E. Moore offered a bold prediction: Within a few generations of development, a silicon chip could be made to hold a billion transistors. He was right. And along with the exponential growth in computing power over the ensuing decades has come an exponential growth in the demand for data — by customers, shareholders and regulators. Behold the data center. With data centers becoming mission-critical facilities, facility executives need practical planning and design strategies to prevent, detect and suppress fire.

What fuels fires in data centers? An analysis of historical data maintained by the telecommunications industry suggests several common fuel sources:

Most facility executives are aware of these common risks and take pains to reduce or eliminate them. However, there are also risks associated with aging infrastructure and improperly executed rack expansions and building modifications: electrical cord failures and shorts and overloaded circuits and improper or poorly maintained terminations.

Of course, aging equipment can corrode over time, creating fire hazards. But these risks also reflect lack of planning for growth. Studies by microprocessor and server manufacturers have found that with an increase in computer equipment load density — from 50 watts per square foot to 100, 200, or even 300 watts per square foot — there is a parallel growth curve in the electrical and mechanical infrastructure required. Although such expansions do not necessarily overtax the fire protection system, which is designed based on an area or volume calculation, they do require a well-planned and well-executed expansion of the electrical and mechanical infrastructure to avoid the potential heat build-up that creates a fire risk.

Planning and Design Strategies

  1. Determine uptime requirement. In planning and designing a fire protection system for a mission-critical data center, the first priority is obviously maintenance of business continuity. The facility owner must determine the level of “uptime” or data availability required. Some corporations have developed tier ratings that drive systemwide design goals; others rely on tier ratings developed by industry organizations such as the Uptime Institute; still others may tap the experience and benchmarks developed by their engineering consultants. The reliability level — whether four nines (99.99 percent), three nines (99.9 percent), or lower — will be a major determinant of the design and costs of the fire protection system, as well as the mechanical, electrical and plumbing systems.
  2. Locate the facility in a low-risk geographic area and a short distance from a fire station. Site selection is critically important from the standpoint of fire prevention, detection and suppression. Avoid highly active geologic and weather zones — for example, earthquake-prone areas or locations where hurricanes and tornados are common. Avoid locations with extreme conditions, such as those prone to wildfires. Avoiding these types of locations will help reduce direct fire risks and those associated with secondary events, such as gas-line breaks, as well as the risks associated with power failures.

    Another consideration is obvious but important: The closer a data center is to a fire department station, the faster the fire station can respond.
  3. Separate support areas from the data floor. Space planning is another important component of a cost-effective fire protection system. Ideally, the data center will be a maximum of two stories high. Support areas, including those that will be staffed, should be physically separated from the data center itself by fire-rated walls and ceilings to allow for installation of the most cost-effective fire protection system, one that limits an expensive clean agent or inert gas suppression system to the data room. Effective space planning also allows for future expansion with the least disruption to ongoing operations.
  4. Detect and notify operators of any fire event as quickly as possible. Most commercial buildings rely on a passive fire detection system that uses ceiling-mounted smoke detectors to sense the products of combustion. Critical facilities need an aspirating high sensitivity smoke detection system that actively captures air from various points in the space and sends it through the smoke detection system. This notifies building operators before the suppression system goes off, giving them time to locate the source of the smoke and determine whether the event warrants release of the fire suppression system and notification of the fire department, or whether they can handle it by powering off a piece of equipment and using a hand-held fire extinguisher at that location. Very early detection can reduce down-time, the damage from a release, and the cost of re-arming the suppression system.
  5. Minimize damage in a fire event due to suppression system release. Due to the risks associated with the release of water above computer equipment, data centers clearly require a different approach to fire protection than the standard wet-pipe sprinkler system used in commercial buildings. A pre-action sprinkler system is one effective alternative. The pipe remains dry until the fire detection system activates a control valve (located outside the data center to avoid damage from leaks), filling it with water. Nevertheless, a release could lead to an emergency power-off situation — and the associated hard crash — and result in significant water damage and clean-up time. Moreover, sprinkler systems require a relatively high heat level (150 degrees F or above) to activate and discharge. A fire that begins in a cabinet, for example, can easily spread before it is detected and suppressed by a sprinkler system.

    A clean agent or inert gas suppression system could be used as a supplemental line of defense for a critical facility, especially in a low-staff or high-risk operating environment. These systems when controlled by an interface with a high sensitivity smoke detection system, suppress fires without damaging IT equipment, and allow staff to get the facility up and running faster. The downside to these systems is their expense, not only in first cost, but also in rearming the system after a release.
  6. Invest in third-party commissioning. Fire-protection systems are complex and expensive, and there is no tolerance in a critical facility for sub-par performance. Third-party commissioning is well worth the investment to ensure that the system has been properly designed and installed and is operating to performance specifications.
  7. Invest in staff training and develop a work authorization plan. Some facilities have 24/7 technical personnel on site, who must be properly trained to respond to a fire event, especially when employing a high sensitivity smoke detection system that allows for discretionary intervention. However, many facilities do not have staff on-site around the clock. Some companies that have maintenance performed by third-party contractors have had a clean agent system accidentally triggered because a worker was smoking. Proactive companies develop a work authorization plan, which outlines every step of each maintenance activity.

An Integrated Design Approach
The new generation of high-density data centers presents three special challenges or risks. In-rack cooling requires more cooling piping infrastructure, which may conflict with other equipment. Higher power calls for more wiring, resulting in a more constricted underfloor space. And the combination of higher power, more cooling equipment and more equipment per square foot means more opportunities for failure. Under certain conditions, overheating may result, increasing the fire risk. This is one reason why it is important to use an integrated design approach to the mechanical, electrical, plumbing and fire-protection systems for mission-critical facilities.

Computer engineers may debate the future of the microprocessor: Already at a billion-plus, can the number of transistors on a silicon chip continue to double roughly every two years? To facility executives of mission-critical facilities, the debate has as much value as the medieval theological question concerning the number of angels who can stand on the head of a pin. What is needed today is a practical, cost-effective strategy for managing fire risks through a conservative, integrated approach to design, construction and third-party commissioning.

John Eagar, P.E., is a senior associate in the critical facilities group of Syska Hennessy Group, Inc., in its Atlanta, Ga., office.