Data center fire protection is challenging, to put it mildly. Today’s IT infrastructure draws an immense amount of power, which runs through electronic components and converts electricity into computational work and heat. Fires in data centers are typically caused by failures in those electronic components, meaning that the typical data center must be considered a fire hazard in and of itself.
Robust fire detection and suppression systems are critical to protect the health and safety of data center employees, as well as prevent damage to the equipment and technology housed in the facility.
Fast response time is essential to minimizing damage during a fire, which leads some data center owners to install Very Early Smoke Detection Apparatus (VESDA) systems to ensure fires are detected and suppressed quickly. VESDA detects microscopic particles that indicate the presence of a fire and alerts staff of the issue.
In many cases, the operator or facility manager will then confirm whether there is a fire, and shut down power to the affected area. If the fire expands past these safety points, fire suppression systems will deploy as needed to limit the amount of damage to the equipment and the building.
The impacts of a data center fire are far-reaching, and can include employee safety, property damage and lost business due to downtime. It’s critical for data center designers, operators, and even legal teams to understand the different types of fire suppression systems so they can decide which method is the best choice for their facility, with regard to safety, protection, insurance, and code compliance.
Fire protection within a data center can be broken down into three distinct levels:
The goal of this level is to protect human lives and minimize damage to the building itself. Fire protection at this level usually comes from handheld extinguishers, sprinklers, firewalls and fire floors, to slow down the spread.
Room-level protection covers the most ground and is guided by standards set by The National Fire Protection Association (NFPA). For a data center to meet code requirements, it must at least adhere to NFPA 75 - Standard for the Fire Protection of Information Technology Equipment.
Rack-level protection is intended to minimize damage to critical equipment and reduce downtime. It uses fire detection tubing installed within the racks to detect a fire and quickly trigger non-water fire suppression before the room-level sprinklers deploy and cause water damage to valuable equipment.
There are two broad categories of fire suppression systems used frequently in data centers: water-based and gas-based.
Because data centers are filled with electronic equipment, it seems like using water to put out a fire would do as much damage to the facility as the fire itself. But in many cases, water-based suppression systems are required to be installed instead of, or in addition to other types of systems.
There are a couple of different types of water-based fire suppression systems: wet pipe/charged sprinklers and preaction sprinklers.
Charged sprinkler systems are always filled with water and ready to deploy. Sprinkler heads are fitted with metal fusible links that keep the water in the pipe until excessive temperature causes them to melt and open the path for water to flow. The specific temperature at which the links melt is dictated by code.
The main risk of this type of system is that if it malfunctions, there is nothing standing between the electronics and a flood of water. This type of malfunction is most often caused unknowingly by contractors bumping the sprinkler heads with ladders and physically knocking the fusible links loose.
Preaction sprinklers, on the other hand, have built-in fail-safes that put a buffer between the equipment and the water. For example, single-lock preaction systems are triggered by a fire or smoke alarm. When the alarm sounds, water is released into the system, at which point it acts like a charged sprinkler system would. If a sprinkler fails or breaks, no water will be released without the preaction of the fire alarm.
Double-lock systems go a step further, requiring a smoke detector trigger as well as a second point of fire detection, before water is released into the pipes. These types of systems can also be restricted by quadrant, so only the area where a fire occurs will get wet.
Although preaction systems provide a level of protection against accidental water damage, they do require periodic testing, which comes with its own set of risks. To test a preaction system, the pipes must be filled with water, then drained. Test procedures often include blowing compressed air through the drained system, but it is not uncommon for some amount of water to remain - usually collecting around pipe fittings. Corrosion can occur over time at these collection points, such that an unexpected fitting failure could occur during testing, or in an unwanted location at the time of an actual fire event.
There is a third type of fire suppression system that uses water, but it’s often classified as a “special suppression” method, like the gaseous systems discussed below. The mist fire suppression system releases a high-powered, very fine water mist that behaves similarly to clean agent suppressors.
Water mist fire suppression systems use less water than sprinkler systems, which reduces the amount of water damage to the equipment and the building. However, the water mist requires high pressure to disperse, so the facility will need propellants like stored nitrogen or a positive displacement pump for the water mist system to work.
Gaseous fire suppression uses clean agent gas, inert gas or chemicals to extinguish fires quickly and efficiently. The most common gaseous fire suppression agents used today are Ansul INERGEN, a clean agent that combines three inert gases: nitrogen, argon and carbon dioxide; FM-200; and Novec 1230.
Unlike water, these agents won’t damage data center electronics, are electrically nonconductive and actuate quickly to minimize and isolate damage.
Clean agent systems using HFC-227ea and FK-5-1-12 are considered safer for people and the environment than some other options. These systems can be used to protect occupied spaces, though you will need to install a notification system to give people time to leave the area before the system deploys.
Fire suppression systems that use carbon dioxide to replace oxygen can not be used around people, due to the risk of suffocation. Facilities using these systems must be thoroughly ventilated before they can be inspected after a fire.
Gaseous fire suppression systems are commonly used in modular data center environments because there is less worry about people getting out. It’s unlikely that an employee in a 30-foot container wouldn’t notice a fire and head for an exit before the alarm sounds.
It’s also unlikely that there will be service individuals nearby for cleanup, so the clean nature of gaseous fire suppression systems is also a benefit.
Inert gas systems have a pretty big footprint compared to Novec systems. A large quantity of inert gas is needed to replace enough oxygen to extinguish a fire. Novec, on the other hand, is stored as a liquified compressed gas, which requires a much smaller storage area.
So depending on the size of the facility you are protecting, and the amount of space available, storing a sufficient number of gas cylinders, versus Novec cylinders, might factor into your decision-making.
Employee safety, protecting valuable equipment and preventing downtime are key considerations when vetting data center fire suppression systems. Knowing what your options are and which codes, standards and insurance requirements you have to uphold can help you select a fire suppression system that fully meets your data center’s needs.
Learn more about the future of data center design, construction and how vendor agnosticism is allowing facilities to reimagine the construction process in our e-book, The Guide to Open Collaboration and Vendor Agnosticism in Data Center.