0% found this document useful (0 votes)
172 views8 pages

Vesda

Brochure

Uploaded by

epicenterrulez
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
172 views8 pages

Vesda

Brochure

Uploaded by

epicenterrulez
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 8

VESDA

White paper

Using air sampling smoke


detection to protect data
centers from fire

Overview
With an electrical usage 5 to 10 times the normal power requirement of a
standard office, Data Centers require specialist air conditioning to maintain
their ideal operating environment.
Conventional passive detectors are neither designed for high airflow
environments, nor do they provide adequate fire warning for areas housing
equipment and data. Installing an aspirating active smoke detection
system is the smartest way to avoid the incidence of fire.
72% of data center applications experience nine hours of downtime per
year1. 90% of businesses go bankrupt within two years of a significant
failure2
The average cost per hour of downtime for a data center is estimated at
US$6.5 M3.
Of the companies that experience a disaster but have no tested business
recovery plans in place, only one in ten are still in business two years later.
The biggest risk to continuous operation within a computer room after a fire
is the smoke damage to electrical equipment, not the flames.
This paper discusses smoke detection systems and their role in prevention
of fire and smoke contamination within a data center facility.
VESDA White Paper: Using air sampling smoke detection to protect data centers from fire

The fire risk within todays Data Center detection chamber and trigger the alarm to alert the
occupants that a fire has commenced. In contrast, if a
Todays computing technology is becoming smaller and computer terminal within the same room had a fault
therefore requires less space, but the heat being within the electronics resulting in a thermal event, it
dissipated by the digital hardware is also increasing. may smoulder for hours before a flame ignites. We
The result is that the heat density on the chip and in the refer to the smouldering stage as the incipient stage to
cabinet is growing at an unprecedented rate. By a fire. During this incipient stage the human eye will not
illustration: see the particles but the human nose may smell them.
The average Intel 486 CPU consumes about EWSD are not sensitive enough to detect smoke at the
10 W, the latest Pentium 4 consumes 100 W. incipient stage of an electrical-type fire. Only a VEWSD
will detect an incipient fire and thus the term VERY
With the processing density and power consumption of EARLY WARNING. This stage of a fire could last for
blade servers it is not uncommon for standard 47 U hours or even days.
cabinets to consume in excess of 21 KW.thats a lot
of heat!! Spot type smoke detectors are passive detectors in
that they wait for smoke and rely on the airflow to
This high heat load requires significant cooling via the transport the smoke to the detector. Therefore their
computer room air conditioning (CRAC) system to performance is affected by high airflow. Since the rate
remove the heat generated within equipment cabinets. of smoke generation in a smouldering fire is relatively
Failure to cool this equipment will result in equipment small, and the airflow velocity in the room is quite high,
over-heating and provide the potential for a fire. the movement of smoke is dominated by the airflow of
Mechanical cooling and airflow movement is an the mechanical systems. Furthermore the smoke
essential parameter within the fire detection design and generated during the incipient stage is not hot therefore
is discussed further in this paper. there is very little thermal lift. This often prevents
smoke from moving directly to the ceiling, where spot
The Detection Strategy type detectors are located, causing the smoke to
Within a data center the type of smoke generated and dissipate more widely. An air-sampling smoke
the dynamics of the airflow creates a challenge for the detection system is active, constantly sampling the air
fire engineer to design an effective fire detection from multiple points throughout the environment. It is
system. It is the detection of smoke that is the most not totally dependant on thermal energy to transport
critical part of the fire protection system. Detection the smoke to the detector.
systems serve the basic function of alerting occupants
within the building of a fire and are used to activate The effects of smoke contamination
other systems such as mechanical exhaust and fire So why is the detection of smoke at the earliest
suppression systems. possible stage important? Because the biggest risk to
The traditional smoke detectors known as Early the continuous operation within a computer room
Warning Smoke Detectors (EWSD) or conventional facility is the smoke damage to electrical equipment,
spot type detectors are of ionization or photoelectric not the fire. In fact according to the USA Federal
type. Ionization type detectors were designed to detect Commission of Communications, 95% of all damage
very small particles such as the type produced by within these facilities is non-thermal.
flammable liquids. Photoelectric detectors detect larger The by-products of smoke from PVC and digital circuit
particles such as those produced by materials like boards are gases such as HCL and these gases will
plastics. Given this fact photoelectric detectors are cause corrosion of IT equipment. Graph 1 depicts the
more suitable to detect the fire type we expect within a increased risk of failure possible with an increase of
computer facility, however there are other factors particulates in a computer room. Even at 16
contributing to photoelectric detectors deficiency within micrograms per square centimeter there is moderate
these environments. corrosion with long-term effects on electronics, at 30
Within the fire industry detectors are categorized as micrograms/cm2 the corrosion is active and the effects
Early Warning Smoke Detection (EWSD) and are short term. Above this the damage to equipment is
Very Early Warning Smoke Detection (VEWSD). In fact detrimental to ongoing performance.
some people use these terms very loosely and do not
differentiate the two correctly. An EWSD system
provides detection of a fire condition prior to the time
that it becomes threatening to the occupants of a
building. Generally this is the time that smoke is visible.
Lets use the example of a paper basket fire within a
standard office. Seconds after the paper has ignited,
smoke will generate and rise to the ceiling. This visible
and hot smoke will eventually enter the smoke

Page 2 of 8
VESDA White Paper: Using air sampling smoke detection to protect data centers from fire

Accordingly, the entire zone is monitored for the early


symptoms of overheating materials, possibly hours
before a fire develops. This generally allows plenty of
time for human intervention, or automatic intervention
by the operation of an electric circuit breaker for
example (which removes the source of heatthe
electric current). The primary role of air-sampling
smoke detection is, therefore, fire prevention.

Graph 1: Contamination and failure probability4


Air-sampling Smoke Detection how it
works
Graph 2: Smoke Density versus Time
Air-sampling smoke detection systems are quite
different from conventional spot type smoke detectors. Graph 2 illustrates the stage at which a VESDA smoke
Aspirating systems typically comprise a number of detector can detect smoke. One of the most exciting
small-bore pipes laid out above or below a ceiling in features of the VESDA System is its flexibility in the
parallel runs, some meters apart. Small holes, also setting of its sensitivity. The detector alarm thresholds
some meters apart, are drilled into each pipe to form a can be set up to 20% obscuration/m. Obscuration is the
matrix of holes (sampling points), providing an even effect that smoke has on reducing visibility. Higher
distribution across the ceiling. Air or smoke is drawn concentrations of smoke, result in higher obscuration
into the pipework through the holes and onward to a levels, lowering visibility.
very sensitive smoke detector mounted nearby, using
The first three thresholds would typically be set with
the negative pressure of an aspirator (air pump).
Alert 0.03 % obscuration/m, Action at 0.06 %
obscuration/m and Fire1 at 0.12% obscuration/m in a
relatively clean environment. Then there is the
opportunity to set Fire2 threshold at
10 % obscuration/m for example, acting as
confirmation of a serious fire event, with the option to
activate a suppression system at that point.
The provision of these alarm thresholds allows for
activating an early and controlled response. For
example, the Alert Alarm (the first alarm) condition may
be used to call local staff to investigate an abnormal
condition. Should the smoke condition continue to
increase the Action threshold may be used to initiate
smoke control, begin a warning sequence via the
evacuation system and alert further staff members via
Illustration 1: Air Sampled through a capillary and paging or SMS to mobile phones. The FIRE1 Alarm
sample point (the third threshold) indicates that a fire condition is
very close or has started. At this stage the building is
The VESDA air-sampling smoke detector is a form of evacuated, the zone on the fire alarm control panel is
air pollution monitor. It has sensitivity some hundreds activated and the signal transmitted to the local
of times higher than conventional smoke detectors, yet monitoring company and fire brigade. The FIRE2 Alarm
its false alarm rate is exceptionally low (according to threshold will activate once the level of smoke is
independent surveys). This reliability comes from its significant enough to calculate that a fire has started
high immunity to the major sources of false alarms and therefore suppression systems can be activated.
dust, draughts and electrical interference.

Page 3 of 8
VESDA White Paper: Using air sampling smoke detection to protect data centers from fire

For the first time, one product can provide very early size to be detected must clearly be less than or equal
warning as well as initiate suppression at a much later to 1.0 kW within a data center if we are to measure the
stage. Of course, if building fire systems and performance of a VEWSD.
procedures have operated correctly, then early
Current testing practice within telecommunications and
intervention should preclude operation of the FIRE2
computer rooms today use a practical onsite test to
thresholdbut its a safety net providing control of the determine the effectiveness of a fire detection system.
last line of defence. In the past, system testing was conducted with a can of
How much smoke should we detect? smoke that was sprayed into the end of the pipe
network or into the point detector to determine if the
Obscuration as a unit of measurement has become the system was working. But this test does not check the
standard definition of smoke detector sensitivity used in systems performance to a real small fire scenario,
the industry today. Obscuration is the effect that smoke which is the benchmark for VEWSD.
has on reducing visibility. Higher concentrations of
smoke, result in higher obscuration levels, lowering A common test used today is the BS6266 "Code of
visibility. Practice for Fire Protection for Electronic Data
Processing Installations". This test involves electrically
Typical smoke detection sensitivities for smoke overloading a short length (1 or 2 meter) of PVC-
detectors: coated wire which produces a small amount of light
Photoelectric: 2%12% obscuration per meter grey smoke barely visible and simulates a smouldering
fire of approximately 100 W.
Beam: 10%25% obscuration per meter
Typically, the test is performed within the room during
Air sampling: 0.005%20% obscuration per meter the commissioning process and the VEWSD should
give an alarm indication within 60120 seconds.
Tests performed by Xtralis have shown that by burning
a measured length of wire within a controlled room we Computational Fluid Dynamic Modelling (CFDM) is
can determine the resulting obscuration/m. For used to determine the effects of such fires within high
example, in a room with a volume of 350 cubic meters airflow environments. It can calculate the theoretical
(3500 sq. ft) burning the insulation from approximately growth of incipient fires, smoke development and the
a 1 m (3 ft) length of 18 AWG wire would produce 0.1% contamination that results from such smoke. Such
obscuration per meter at ceiling level, easily detectable models can be used to determine the level of
by a VESDA very early warning smoke detection contamination (mass of particulate per cubic meter ) for
system. Obviously, having 1 meter of wire burn is a specific fire sizes. This is useful in estimating the
significant fire event in a Telco facility. amount of contamination to which IT equipment is
exposed during various fire conditions. As shown in
Graph 1 the contamination exposure caused by fires
will increase the probability of IT equipment failure.
Use of a VEWSD detector to detect such contaminants
can reduce the risk of such damage occurring. In many
cases the contaminants are present at very low levels,
often as a result of high background smog/smoke
levels, often introduced from poor quality air during the
use of economy cycle HVAC. Without the use of
VEWSD the low levels of these contaminants can go
un-noticed for long periods of time causing insidious
but permanent damage.
In addition, the use of the event log of a sophisticated
VEWSD can be used to support a warranty or product
replacement claim on an equipment vendor where
Graph 3: Smoke obscuration measured based on a
equipment fails within its warranted terms of use. This
burning wire within a 1000 m2 (10000 sq. ft) room
is especially important where the vendors warranted
The amount and color of smoke created in a computer terms of use reference the quality of the environment
room during a fire is dependent on the type and rather than the deposition of contaminants on the
amount of material burned. Smouldering combustion of equipment, as is often the case.
a printed circuit board may produce a heat release rate
of one or two kilowatts and the heat release rate of a
single resister is as low as 10 W. By comparison the
heat released from a paper basket fire may be between
2 to 4 kW (UL standard paper burn (3 sheets)). The fire

Page 4 of 8
VESDA White Paper: Using air sampling smoke detection to protect data centers from fire

Beyond conventional design techniques Illustration 3 shows the ASD pipe network configured
for both circumstances; where the CRAC is operational
Although the design of fire protection systems has and out of service. The sampling pipes on the ceiling
primarily been based on traditional prescriptive fire and within the floor void are used for detection where
codes, there is an increasing emphasis on the CRAC is out of service. The pipe used to detect
performance-based codes that address individual smoke across the return air path is used for detection
environmental requirements. Performance-based where the CRAC is operational. This design method is
design determines the best fire protection system by suitable for rooms where the ceiling height is not tall
assessing the function, risk factors, and internal and room is small in area.
configuration and conditions of a specific environment.
When designing a fire detection system for VERY
EARLY WARNING the designer must consider the
following:
1. The airflow characteristics and the air change rate
within the room.
2. The coverage area per detector or sample point.
3. The sensitivity required per sampling point.
4. The room size and characteristicsraised floor, tall
ceilings etc.
5. The annunciation of emergency response systems.
6. The activation of mechanical control systems such
as air extraction and suppression systems.
The detection system must be designed for conditions
when the air handling system is either operational or Illustration 3: ASD pipe network configuration that
out of service. provides smoke detection when the CRAC is
operational and when it is not.
Illustration 2 shows the detection method on the CRAC
return air path, relevant in circumstances where the For large rooms with high airflows it is recommended
CRAC is operational. This method of VEWSD is that a combination of on-ceiling detection, underfloor
suitable for rooms that use EWSD spot type detectors detection and return air be used.
as the detection scheme in circumstances where the
CRAC is out of service. Coverage area
The area coverage of the detector is a very important
criterion of the design. This is true from both a
performance and cost-effectiveness perspective.
Illustration 4 shows a grid layout for an ASD detector
for a 2000 m2 (20000 sq. ft) area (this is the maximum
area coverage permissible within the BS, AS and
NFPA codes). Each sample point of an ASD detector is
treated the same as a spot type detector within most
prescriptive codes. You can see below that the area
coverage for a sample point is effectively the circle or
close to the square around it, which is 10 m x 10 m =
100 m2 (10000 sq. ft) (illustration 4 is designed as per
Australian Standard 1670 and would be suitable for a
low airflow environment). For ASD applications in high
airflow environments, we can decrease the area
coverage for the sample point by adding more holes
and making the distance between each pipe less.

Illustration 2: Smoke detection at the CRAC return air


grille.

Page 5 of 8
VESDA White Paper: Using air sampling smoke detection to protect data centers from fire

5m 10m 10m 10m 10m 5m

5m
Area

Coverage
10m

ASD 10m
Detector

Illustration 4: Grid layout


for an ASD detector for a 10m
2000 m2 (20000 sq. ft)
area.
5m

Sampling points
Sampling pipes

The prescriptive codes and standards today describe Section 5.2.5.2 Point DetectorsRecommended area
detection techniques for on-ceiling detection. But new coverage per detector for the different location zones
codes such as NFPA 76 "Standard for the Protection of are given in table 1. As follows: EDP equipment room
Telecommunication Facilities" is the first code that uses (ceiling height above 3 meters), require a maximum
a prescriptive and performance based approach for the spacing of 15 - 25 square meters (150 - 250 square
fire protection of telecommunication facilities. This code feet).
specifies both the area coverage as well as the
As well as codes there are insurance companies such
sensitivity of the detector. Presently NFPA 76 requires
as Factory Mutual who specify in their Property Loss
that Every type of sensor and port installed in a space
Prevention Data Sheet (5-48) for Automatic Fire
shall be limited to a maximum coverage area of 200 sq.
Detectors page 7 "A maximum coverage of 200 sq. ft.
ft. (reference page 51, section 6-5.3.1.2*) Exception:
(20 m2) per detector may be necessary where room air
When (2) levels (high and low) of ports or sensors are
is changing at a rate of 20 air changes per hour
provided, each level shall be limited to a coverage of
400 sq. ft. or less per port or sensor. So the message is that more detectors should be used
in high airflow environments to increase the chance of
NFPA 72 National Fire Code for the USA
seeing a fire, however this requirement can be offset by
recommends the area coverage for spot type detectors
the use of a VEWSD which can support a large number
to be reduced within high airflow environments to as
of sampling points in a single pipe network.
low as 11.5 m2 per detector for rooms that have 60 air
changes per hour. Sensitivity of Air-Sampling Smoke
British Standard BS 6266 (1992), Section 5.2.5.1 Detection
Detector SpacingGeneral From the point of view of
Although reduced spacing will increase the probability
automatic fire detection, EDP areas present fire risks
of smoke being detected, it does not determine if the
quite different from those in many other premises. The
smoke generated has an obscuration density high
concentration of high value equipment, sensitive to
enough to trigger an alarm. Therefore the sensitivity of
damage by even a small fire or smoke, and particularly
the system is also fundamental to the design of the
the high potential consequential losses, make it
VEWSD system.
important to use close spacing of detectors. Detector
density should be high enough to enable the smallest The sensitivity of the aspirating detection systems
fire to be detected quickly without unduly increasing the sampling point is extremely important to ensure
false alarm risk. A reason for a higher than normal consistent and sensitive detection within the zoned
density of detectors is the influence of the air- area. But what codes and standards do not take into
conditioning system, which dilutes the smoke being account for air-sampling smoke detection systems is
produced by fire. their ability to use cumulative air sampling within an
environment.

Page 6 of 8
VESDA White Paper: Using air sampling smoke detection to protect data centers from fire

Cumulative air sampling refers to the way the air- iii. The detection occurs before any spread of the risk;
sampling smoke detector samples smoke over the loss can be minimized:
network of sampling points, allowing each to contribute Smoke is not allowed to contaminate or
to the smoke being sampled at the detector. Within a otherwise affect other systems in the data
high airflow environment this phenomena becomes centre
very useful as particles of smoke are spread through
the room allowing the cumulative sampling effect to Compartmentalization ensures that in worst
take place. cases the estimated and possible maximum
loss and business interruption estimates are
Take the example of a 200 square meter room with 10 minimized (for insurance assessment)
sample points on the ceiling. If the detector sensitivity
is set to 0.1% obscuration/m this effectively makes iv. The background dust and smoke levels within
each sample points sensitivity 0.1 x 10 = 1.0% sealed enclosures are relatively consistent. Also,
obscuration/m. That is, if only one sample point was the airflow dynamics within a sealed enclosure can
exposed to smoke it would require 1.0% obscuration/m be predicted with relative confidence by
to trigger an alarm. This is because the fluid mechanics computational fluid dynamic models. This ensures
of the model takes into account dilution caused by the that detection systems can be designed, built and
other holes. commissioned with confidence of their efficiency
and performance.
Using the same example, if smoke enters three holes
v. Fire responses can be more automated, and cost
the effective sensitivity required to trigger an alarm is
and downtime from fire responses, such as use of
0.1 x 10 divided by 3 = 0.33% obscuration/m. Clearly,
suppression, is reduced.
cumulative sampling allows much lower levels of
smoke to be detected and therefore, allows very early vi. Better control of the issue management and
warning. escalation processes is possiblean alarm can be
routed to the data centre manager as an
If the same room was designed with EWSD and each "environmental alarm", rather than reporting via the
detector was rated at 5% obscuration/m, the alarm main fire alarm system. This staged response to a
would only trigger once the smoke density has reached fire threat allows IT staff investigation and possible
this point throughout the room or at one detector. intervention, an ability to move processes or data
In-cabinet and integrated-equipment from problem equipment, action such as power-
down of problem equipment and, if necessary,
detection suppression of an escalated fire. Such a staged
Interest is developing regarding the application of ASD response will often negate the need for
within data racks & enclosed equipment cabinets and suppression to be fitted or, if fitted, will negate the
integrated in specific equipment or assets. It is need for expensive suppressant to be released.
desirable to fit ASD within these cabinets because in vii. Cause & effect is localised, ie. fire controls used
some circumstances it would not be acceptable for are specific to the cabinet rather than the room.
smoke from a fire within the cabinet to breach the Use of common area protection systems means
cabinet, enter the data centre facility, contaminate that the common area is unprotected until the
other systems or processes and possibly activate main system is re-charged.
alarms and suppression systems.
In-cabinet smoke detection and action enables an viii. Integration with existing communications systems
excellent very early warning solution because: is possibleremote and centralized monitoring and
i. The sampling is performed closest to the source of maintenance (especially for unmanned or
the fire, before dilution, which allows earliest automated facilities), eg over LAN, MAN, WAN
detection becomes cost effective. Also, coupling advanced
smoke detection with a full suite of environmental
ii. Sampling within the enclosure allows clear monitoring systems (power loss, access-control,
identification of the source of the problem. This security, temperature, water loss, humidity etc)
addressability reduces time, effort and error in offers a number of synergies.
identifying and remedying the problem.

Page 7 of 8
VESDA White Paper: Using air sampling smoke detection to protect data centers from fire

Conclusion References
Due to the huge financial loss and potential business 1. Standish Group Research, 1998
risk, a data center facility cannot risk downtime 2. Christensen, Steven R. and Schkade, Lawrence L.
especially of the size and duration potentially caused "Financial and Functional Impact of Computer
by fire and smoke contamination. The most important Outages on Business," University of Texas at
system that contributes to the prevention of fire and Arlington
smoke damage is a very early warning smoke 3. Computerworld, August 4, 1997
detection system such as a VESDA system, that meets
4. 2nd International Fire Protection Seminar Buenos
the performance objective to detect smoke at the very
Aires 1999 - Telecommunication Facilities.
early stages of a fire.
5. BS 6266 (2002)
The VESDA Air-Sampling Smoke Detection System
6. NFPA 72 (2002), NFPA 76 (2002)
features provide the designer flexibility by meeting
design requirements of prescriptive codes as well as 7. AS1670 (2004)
facilitating use of todays performance-based fire
engineering methodologies. These enabling features
include:
Detection of both small incipient smouldering fires
and large flaming fires
Flexibility to design on ceiling, under floor voids,
cable ducts and across return air intakes, as well
as in targeted equipment sampling
Multiple alarm levels that can be used to provide:
o initiation of orderly shutdown of computer
systems and processes and power systems
o removal of contaminated air (via activation of
air handling systems, baffles etc)
o communication of reliable early warning (to fire
wardens, brigades, etc.)
o initiation of staged evacuation
o initiation of automatic suppression

www.xtralis.com
The Americas +1 781 740 2223 Asia +852 2297 2438 Australia and New Zealand +61 3 9936 7000
Continental Europe +41 55 285 99 99 UK and the Middle East +44 1442 242 330
The contents of this document is provided on an as is basis. No representation or warranty (either express or implied) is made as to the
completeness, accuracy or reliability of the contents of this document. The manufacturer reserves the right to change designs or specifications
without obligation and without further notice. Except as otherwise provided, all warranties, express or implied, including without limitation any
implied warranties of merchantability and fitness for a particular purpose are expressly excluded.
This document includes registered and unregistered trademarks. All trademarks displayed are the trademarks of their respective owners.
Your use of this document does not constitute or create a licence or any other right to use the name and/or trademark and/or label.
This document is subject to copyright owned by Xtralis AG (Xtralis). You agree not to copy, communicate to the public, adapt, distribute, transfer,
sell, modify or publish any contents of this document without the express prior written consent of Xtralis.

Document: 11781_01
Page 8 of 8

You might also like