A Guide to Datacentre and Server Room Integrity Testing

Fires in a server room or datacentre are a risk that can be mitigated against by installing a suitable suppression system and testing the room’s integrity. The aim of the testing is to ensure that a sufficient level of the fire suppression agent or gas is retained for a long enough period to extinguish the fire and prevent reignition. Fire suppression systems and their components are normally subject to annual maintenance by a certified fire engineer. The same should apply to room integrity testing and some organisations may be obliged to do so.

Typical Causes of Datacentre Fires

Datacentre fires can result from a range of factors. Within the server room space, rack power densities continue to increase and it is not uncommon to find server racks drawing power from 5kW up to 30kW for high power density servers. A sudden loss of air conditioning can immediately lead to a critical ‘hot-spot’ and over a short timeframe, fire within a server rack.

More common causes include poor electrical distribution system design and maintenance. Electrical faults and conductor overheating can result from poor specification, overloading or harmonics within the electrical system.

Other causes of datacentre fires include rodents biting through exposed cables (Yahoo, Santa Clara), lighting strikes (Microsoft, Dublin) and in Welwyn Garden City, an explosion of poorly installed fire suppression gas cannisters at an HSBC datacentre. Other notable datacentre outages include BT, Belfast which lost service for a day due to a fire outbreak. Most recently in 2019, Wells Fargo experienced a datacentre fire following routine maintenance that resulted in a cascade effect knocking out operations.

Fire suppression systems can also discharge accidentally due to a false alarm trigger, as happened at the Dutch banking company, ING in 2016. This shut down datacentre operations for the bank for almost 24 hours during which clients could not get access to their money. Whilst this was an accidental discharge, the incident shows how important it is to have a reliable fire system in place and one that can avoid triggering from accidental and/or false alarms.

In the future, a greater adoption of lithium-ion batteries, whether used in uninterruptible power supplies or energy storage systems could increase the potential for fire within the IT space. Lithium is highly flammable when exposed to air. The use of lithium batteries in a server room or datacentre environment will increase the need for robust fire protection solutions.

For more information on datacentre fire reports visit:
https://www.theregister.co.uk/2019/02/07/wells_fargo_outage/

Thermal Management Systems

Thermal management is one of the most critical aspects for any datacentre or server room operator to manage. From a cooling point of view, the more efficient the cooling system, the higher the energy efficiency and the greater the guarantee of providing an ambient and humidity-controlled environment.

Whilst a cooling system failure may not automatically lead to a fire, it can be a potential cause. For this reason the larger the facility, the more redundancy is built into the air conditioning design whether the core system is designed around traditional computer room air conditioners, free-air or liquid cooling.

Whilst a fire suppression solution is sometimes seen as a secondary system for some installations, it is as important a consideration. The effectiveness of the fire solution is also directly affected by system design, suppression agent/gas selection and the integrity of the room being protected. Where a room is not adequately sealed, any fire suppression agent released can dissipate too quickly to effectively suppress the fire and lower its temperature below a reignition level. Poor sealing can also lead to structural damage to the server room and building due to over-pressurisation as te suppression agent/gas is released.

Room Integrity Standards

Room integrity tests are designed to ensure that the extinguishing agent released from a fire suppression solution can be held within the room at specific heights for a period of 10 minutes or more to prevent reignition of a fire. An initial integrity test should be carried out as part of the commissioning of a new fire suppression solution and then as recommended by BS ISO 14520 and 15004, annually as part of planned maintenance schedule.

Aside from providing ‘peace of mind’ that the fire suppression system remains a viable means to protect the room or enclosure, the testing and certification may also be required for insurance purposes.

The reason for annual testing is that the structural integrity of a room changes over time, affecting the potential for leakage. A room’s integrity can change due to general ‘wear and tear’, structural ageing, weather-related environmental factors i.e. high or low temperatures and work in site including cabling and other works.

Understanding Gas Leakage

For a fire suppression system to operate effectively, the correct quantity of gas agent needs to be discharged and at concentrated levels to extinguish the fire. The Fire Industry Association (FIA) provides a document entitled: ‘Guidance on Room Integrity Testing and its Interpretation’.

When a fire suppression gas is discharged into a room or enclosure, the gas mixes with air within the room. Some of the mixture is expelled from the room or enclosed space, due to over-pressurisation. Once discharged the gas to air concentration must be 30% greater than that needed to extinguish the fire and with a retention period long enough to allow the fire to cool below its reignition level (referred to as the ‘Hold Time’). The fire suppression gas is heavier than air and results in a ‘Column Pressure’ which refers to the pressure at the bottom of the enclosure space due to the heavier than air elements. This means that the gas/air mixture can leak through lower vent paths and poorly sealed boundaries. Air can also leak into the upper space within the enclosed space to replace the escaping air volume.

For more information visit:
https://www.fia.uk.com/uploads/assets/uploaded/ece4931c-46c0-44f7-88130ab7d78565d9.pdf

Server Room Integrity Test Methodology

Door fan test methodology is the preferred way to test the integrity of a room installed with a fire suppression system. The testing provides a low-cost way to predict retention time without releasing the extinguishing agent from the fire suppression system. The practice meets the general requirements specified in BS 5306, BS:ISO 14520, NFPA 12A, NFPA 2001 and the BFPSA Code of Practice for Gaseous Extinguishing Systems.

Door Fan Testing Procedure

The goal of using a door fan is to be able to measure pressure differentials between the enclosed room and outer areas. The door fan is placed in the doorway to the server room or datacentre and airflow and pressure measurements are recorded. From this data, the leakage characteristics of the room can be calculated including the predicted retention time. The general door fan test procedure includes the following steps:

1. Site Survey: a survey may be necessary in order for the the size of the project to be scoped out, including room and system characteristics, site restrictions, permits and health & safety requirements. This may only be necessary where this information cannot be obtained via telephone and email.
2. Measurements and Data Recording: the room or enclosure is measured (including the maximum height) in order to generate a planned layout, noting the type and size of fire suppression system and quantity of clean agent.
3. Single Space Identification: where required, doors within the enclosed space are opened and false floor and ceiling tiles removed to create a single test space. A return air path is created outside the enclosed space using opening doorways and windows as required.
4. Door Fan Placement: the door fan test kit is placed within the doorframe leading into the room to be tested. The room can remain in use and people may enter and leave except when data is being taken. The number and type of door fan(s) used to pressurise and depressurise the enclosed space to the require column pressure and the fan pressure required are noted for the computer modelling stage.
5. Air Handling Systems: if the enclosed space has air handlers (for input supply or extraction), these will need to be set for normal operations as would occur when the suppression system would discharge. This could include dampers closed and fans in standby or off-mode. This is only for the duration of the pressurisation test. Air conditioning units can be left on during the test to maintain the ambient temperature. The overall configuration is recorded for the final report.
6. Computer Modelling: the data from the tests is entered into a specialised software program which is designed to model and calculate the design concentration and column pressures that would be seen when the clean agent/gas is discharged. Airflows, leakage area(s) and retention times are generated. If the retention time is greater than 10 minutes the room is deemed to have passed the integrity test. If the results are lower than expected or required, further investigation can be carried out. This can involve the use of chemical smoke pencils.
7. Chemical Spoke Pencils: may be used in addition to the door fan(s) to help to identify leaks. The pencils produce a small amount of smoke and are only used at the perimeter to the enclosed space and away from servers and other sensitive electronic devices. Identified leaks can be sealed temporarily and the integrity test(s) repeated.
8. Final Test Report: a complete report including findings and recommendations is generated following the test(s). Details included cover the room(s) and enclosed space(s), fire suppression system design and components, pressurisation results, a graph of the predicted retention time and identification of leakage areas and any others that require corrective actions to be taken.

Summary

For server rooms and datacentre protected by fire suppression systems, room integrity testing should be a regular and annual event, carried out as part of a planned maintenance schedule. Without this, the resilience of a fire protection system can weaken over time and lead to instances where the release of a fire suppressant extinguishes a fire but fails to prevent reignition.