FIRE TEST ON COMPUTER TAPES

The results of this test were published in many fire and safety journals and resulted in redesign of computer media stores, video tape stores - and even ships. Although these tests were undertaken many years ago, the principles and science behind them remains good today. Despite the publicity surrounding the tests at the time, we have been surprised at how few organisations are still unaware of the potential danger and impact of fires involving magnetic media.

Background to the test

Tests were undertaken at Harwell Laboratory in 1988 on British Standard optical smoke detectors in computing environments.  The installations had been considered to be protected to British Standard^* and had been tested as operational.  The Harwell tests, using a Roscoe smoke generator, showed that, in some machine rooms, air turbulence from computer equipment, peripheral devices and air conditioning rendered these optical detectors ineffective.   This has serious implications for the security of computer installations protected only by optical fire detection and ramifications for all similarly protected areas with air conditioning and equipment fitted with cooling fans.  These tests led to upgrade of fire detection in the Harwell Computing Centre (and to a view that BS6266 (1982) required amplification).  Harwell had already identified computer tapes as a possible fire hazard and the detector tests emphasised the need for further investigation of computer tape flammability and storage.  

Some tests involving the burning of computer tapes had been held at Harwell in July 1985, and it was decided to build on the findings of these with a burn of 1,000 open-reel computer tapes.   Tests were supervised jointly by the Chief Fire Officer of the Harwell Laboratory Fire Brigade and Harwell Computing Centre staff.  A further series of tests was therefore planned to establish the flammability of computer tapes.

Two tests took place in the breathing apparatus chamber of the Fire Station of Harwell Laboratories and one was conducted in a disused building.   This report relates mainly to that last and largest test, carried out on the 27 September 1988.   A video record of this test was kept.

The building used for the test was approximately 40' x 40' with a ceiling height of approximately 15'.   The walls were of 9" brick on a solid concrete base and fitted with a solid 9" reinforced concrete flat roof.   The building was provided with high level windows on three sides; one 4' door opening; and two 4' low level metal framed windows to the front of the building.  The windows had the glass removed for safety.  The oxygen flow was deemed to be representative of an air conditioned tape store.  Two fire appliances containing a total of 800 gallons of water were standing by.   No hydrant was available.

For the test the building was fitted with three high level storage racks that held approximately 1000 old computer tapes.  It should be noted that the racking was obsolete and expendable - tapes sat within the shelves of two of the three racks.   On the third, tapes were suspended (as they are on most modern racking).

___________________________________________________________________

* BS6266 (1982), the British Standard Code for Electronic Data Processing Installations.   The fire detectors complied with BS5445 (Specification for Components of Automatic Fire Detection Systems)

The Tape Burn

To start the fire a small amount of litter such as might be found in a tape library was placed at the foot of one of the tape storage racks.   The test commenced by lighting the combustible debris with a single match.

The fire took hold very rapidly, quickly engulfing the first of the tape storage racks.   The plastic tape cases melted quickly, and drops (in some cases streams) of burning plastic fell to the floor and formed blazing puddles, spreading the seat of the fire. 

At this time little smoke was evident.   Flames reached the ceiling and the fire spread rapidly through the racks.   However, within 5-6 minutes thick black acrid smoke was emitted.   For about 2-3 minutes this smoke levelled out at approximately 8' above the floor level until the fire completely engulfed the remaining two racks of tapes.   The smoke then built down, after 8-10 minutes from ignition, until only the bottom 12" offered any visibility at all.

The intense heat caused the reinforcing steel in the concrete to expand and pieces of it broke from the ceiling: the concrete then began breaking up rapidly, sounding like machine gun fire.

The fire was allowed to continue until evidence appeared that structural damage to the building was taking place.   A large concrete beam running across the full frontage of the building moved outwards about 1" and smoke could be seen seeping from under it.   Also a steel beam running from front to back of the building and built into the brick walls had expanded and pushed out the brickwork several inches at the rear of the building. From signs of metal expansion and structural damage, the temperature was estimated to be in excess of 1000^oC.

The original intention had been to permit the fire to burn for 27 minutes.   This time was based on Lampertz's assessment of the usual time taken by a County Fire Brigade from outbreak of a fire to attend and start hosing.   This equates to the Home Office Standards for risk Category D, giving an attendance time of 20 minutes from notification to the Fire Brigade - fire detection and notification account for the 7 minute difference.  However, because of the increasing threat to the integrity of the building, after 15 minutes it was decided to extinguish the fire.   This broadly equates to Home Office Standards for attendance at a Category C fire.

Two firemen wearing breathing apparatus and standard fire brigade fire fighting uniform commenced to try and extinguish the fire using a fire service hose reel fog gun.   Such was the intensity of the fire that they were only able to penetrate about five feet into the building.   The hose reel fog gun had little effect except to wash the air of the smoke (which appeared to be partially inhibiting the fire) and thus to permit the deep-seated fire to burn more fiercely.

One could also see the extent of the fire more clearly.   All the tape racks were alight from top to bottom.   The frames were distorting with the heat, but the older-style racks still supported the tapes that sat within them.   The fire however, had burnt through the catches on the suspended tapes and these had fallen to the ground, spreading the base of the fire.  (In the July 1985 test (see photos) tapes had been stored on Dexion racking which, within 7 minutes of the start of fire, had buckled in the heat, tipping tapes onto the floor in a similar way and causing the fire to spread widely).

About 600 gallons of water had by this time been discharged into the fire. One fire appliance had been despatched to refill.

At this point it was decided to try and extinguish the fire using High Expansion Foam produced on Turbex foam generator.   74000 cubic feet of foam was introduced into the building.   Some 66000 cubic feet of foam was destroyed by the fire due to its intensity before the fire was under control (though not totally extinguished).  The application of foam continued until it was safe to enter the building and extinguish pockets of fire with the hose reel fog gun.

The foam and water quiesced the fire, but about eight minutes after it appeared extinguished it again erupted.   Within four minutes it was clear that it would become just as fierce.   This apparent extinguishment and rekindling repeated several times until about an hour and a half after ignition the fire appeared finally extinguished with water.   As in previous tests the tapes (which had welded together as the plastic covers melted) had to be broken up with pick axes before all were satisfied that the fire was finally extinguished.   As the tapes cooled, a reason for the deep-seatedness of the fire became apparent.   When the soft molten plastic tape cases were hit by the cold water or foam, the outer edge of the case solidified.   This solid plastic had formed a shell, repelling further water and protecting the flames burning beneath it. 

Despite the ferocity of the fire over half of the total flammable material remained unconsumed.

EXAMINATION OF COMPUTER TAPES AND THEIR PYROLYSIS PRODUCTS

It was decided to undertake laboratory analysis of tapes and pyrolysis products to establish their constituents.   Both "traditional" open-reel tapes and the more modern 3480 cassette tapes were examined.

Open Reel Tape

Infrared (IR) spectrophotometric examination of the oxide coated tape gave a spectrum which was of poly(ethyleneterephthalate), i.e. Mylar.   Both sides of the tape spool gave IR spectras of polystyrene.   The red portion of the outer retaining ring gave an IR spectrum of a styrenated acrylonitrile-ester polymer, while the spectrum of the white portion was that of ABS plastic (acrylonitrile/butadiene/styrene).

Pyrolysis Products

Pieces of tape plus spool (weight ratio approximately 4:1) were pyrolysed at 600^o in a stream of air flowing at 500ml min^-1 and pyrolysis products were collected in a 10cm path-length gas cell for examination by IR spectrophotometry.

The test was deliberately done at 600^o because that is the temperature at which the most compounds will be found.   At very much higher temperatures (over 1000^oC most toxic chemicals are destroyed.   Also no two fires are the same and even at 1500^o there will be a range of temperatures down to 200^oC or so.

The major components were carbon dioxide, carbon monoxide and water vapour with trace quantities of styrene monomer and a carbonyl compound.   The absorption coefficient of carbon monoxide is very much less than that of the dioxide, so that, although the dioxide absorption bands were greater on one test, there was more monoxide present.

A red piece of the outer retaining ring, treated in a similar way, produced the same major components plus smaller quantities of methane, hydrogen cyanide, acrylonitrile and unsaturated hydrocarbons.

The amount of cyanide could not be determined easily.   It depends on the quantity of cyanide forming material and on the conditions under which it is burnt.  Hydrogen cyanide has a short term (10 minute) exposure limit of 10 ppm in the atmosphere.   The characteristic smell is noticed at about 2 to 5 ppm.

3480 Cassette Tape

The IR spectrum of the oxide coated tape was consistent with that of Mylar while those for the outer case and the spool indicate that these are polycarbonates.

The 3480 tape consisted of Mylar tape as tested for the large tape and a polycarbonate case.   The case did not contain a nitrogen so cannot make cyanides, otherwise product was similar to open reel tapes.

Danger to people from toxicity

The main dangers are not so much the cyanides but from the toxicity of different levels of combustion in varying circumstances causing coughing: this then causes large amounts of 'bad' (i.e. oxygen low, carbon dioxide and monoxide high) air to be inhaled.   This air could also be very hot and cause burning of the respiratory tract.  The long term effects of inhaling the toxic compounds identified is not known.

CORROBORATIVE TESTS

Tests on Halon 1301 versus Water Sprinkler Fire Protection Equipment for Essential Electronic Equipment were undertaken by the Engineering and Services Laboratory, United States Air Force (USAF) Engineering and Services Center, Tyndall Air Force Base, Florida and a Report (ESL-TR-82-88) was produced in 1982.

Primarily these tests were about relative performance of extinguishants and extinguishants' effect on electronic equipment.   One test in particular, however, (Test A-8) tends to corroborate the Harwell experiments.   This was the only test involving magnetic tape.

In the USAF test, detection  equipment in a custom-designed test room measuring 26' 5" x 16' x 7' comprised two 3040RC Series photo-electric smoke detectors and two PID-B ionization sensors mounted on a suspended ceiling and one photo-electric smoke detector and one ionization detector located in the sub-floor space.   The air exchange system was switched off.                               

The USAF test took place on 4,800' (i.e. two reels) of unwound magnetic tape and the Halon system discharged at one minute from start of fire.   The test had to be cut short to preserve video equipment used to record the test.   After a twenty minutes Halon soak, the 25% of material remaining immediately reignited when the facility was re-opened.

It is also worth noting that, in the USAF tests, the force of the Halon discharge blew several ceiling panels out of their ceiling grids.

In the event that the fire was not detected until it was well established, displacement of the ceiling tiles on discharge of Halon would allow the fire into the roof void.

CONCLUSIONS

The intensity of the fire surprised all who witnessed the test - despite having seen earlier, smaller tests.   Had the fire taken place in a normal tape store, with false ceilings and raised floor, it is believed it would have rapidly spread through the roof and sub-floor voids and that smoke would have quickly dissipated into the roof void.  Use of modern suspended-tape racking would probably assist the spread of fire as the plastic retaining tapes burnt through and tapes fell.   Plastic 3480 type cassette tape racks would ignite rapidly, and cassette tapes might be spilt as the plastic racking melted.

Taking into account the earlier tests on optical smoke detection, it is clear that, in some circumstances, a fire in tapes stored in an air conditioned environment - particularly inside an unmanned computer room - could become well established before it was detected.   Once established, it quickly becomes deep-seated.

Where possible tape stores should be at ground floor level in a detached building of solid construction with arrangements built into the walls for the introduction of high expansion foam.  The building should be equipped with Very Early Smoke Detection Equipment (VESDA) or a mix of optical and ionisation smoke detection equipment and a fixed fire fighting installation  of Halon or water sprinkler.   (Given that heat does not damage the tapes, water does not present a recovery problem: Harwell has successfully dried out water-soaked computer tapes which, when tested, gave error-free results - video tapes and music masters have similarly been successfully recovered by Harwell).

Adequate continuous water supplies should be available within approximately 150 feet of the building.

If a detached building is not possible then it should have solid brick/block party walls having a fire resistance of not less than two hours.   The computer store should not have other occupied floors above it.   An exception could be made here if the proposed computer tape store did not have any apertures that could allow fire spread by curling of fire up the face of the building.   The ceiling of a tape store with floors above it should have a fire resistance of not less than two hours.   Access into the tape store should not be through doors directly from an adjoining building.   Access direct from the open air is highly desirable.

Any mechanical ventilation of a tape store should be independent of any ventilation system that serves the building a tape store may be attached to.   Air conditioning should be powered off before extinguishant is activated.

Below ground storage of computer tapes may be acceptable if the following conditions are met:

• the underground storage area is not within a building;
  
• it should be fitted with a smoke detection system coupled to a fixed Halon or water flooding installation;
  
• in order to fight a fire either by flooding with water or foam it should not be necessary for firemen to enter the basement;
  
• persons should not be employed to work in a below ground tape store other than for tape turnover or engineers carrying out maintenance work.

It is suggested that the Fire Officers on the various sites should be informed of all stocks of tapes over, say, 50 in number so that they can make an assessment of the risk and give the necessary advice based on the guidance contained above.

It is possible that where stocks of tapes are relatively small (below 100) they could be stored within buildings in fire resistant cabinets: however, it is accepted that, in a large and busy computing centre, this may not be practicable.

Most fires impacting computer installations start from outside the computer room: therefore, where computer tapes have to be kept within the computer room, they should not be stored adjacent to party walls.

Many computer installations, although having British Standard fire detectors installed, may not be as sensitively protected as they believe.   Given that; the ready flammability of computer magnetic tape; the difficulty in extinguishing magnetic tape fires; the value of the computing installation; and corporate dependence on it, the Harwell tests give rise for concern.

It is clear that organisations which store bulk magnetic media (including video tapes) should, as a matter of urgency review their storage conditions and the detection and extinguishants installed at the stores.