Silent solutions for hard disk drives
Published: 10 July, 2013
When the fire protection industry became aware of potential damage occurring to hard disk drives during discharges of automated dry extinguishing systems, Siemens decided to find an answer to the problem, writes Switzerland-based Juan Jose Merlo Latorre of Siemens, Infrastructure & Cities Sector, Building Technologies Division.
Data centres and server rooms are at the heart of the modern business world. With an estimated 35 million servers worldwide processing ever-increasing quantities of data throughout a global network of several billion devices, the need to ensure business continuity and protect knowledge from loss is essential.
When it comes to a fire in a data centre, seconds count. In a data centre or server room, the hardware, electrical power and extensive cabling that drive computing systems create a constant risk of potential ignition. In addition, many thousands of plastic components supply plentiful sources of combustible material. All these factors lead to an increased risk of fire.
In 2009, the fire protection industry became aware of potential damage occurring to hard disk drives (HDDs) during discharges of automated dry extinguishing systems. The damage ranged from automatic shutdowns to more severe disturbances, with corresponding losses of data. As a long-time pioneer in the development of advanced extinguishing products Siemens decided to find an answer to the problem.
Tests were performed in three steps to verify the underlying reason why the HDDs were disturbed during the discharges.
The three tests were based on the following questions:
1. Is the damage caused by overpressure?
2. Is the damage caused by high noise levels?
3. Are the laboratory results transferable to actual extinguishing tests?
Is the damage caused by overpressure?
The delivery of a large amount of extinguishing agent into a protected area within a short period of time generates overpressure in the room. The first question addressed was whether overpressure was causing the reported damage to the HDDs.
For data centres, inert gas extinguishing systems are strongly recommended because of the high sensitivity of the server room equipment. Inert gases operate by reducing the oxygen volume concentration in the protected area to a level where the fire cannot burn.
Depending on the fire risk, the oxygen volume concentration is generally reduced to between 10% and approximately 14% in order to extinguish the fire. The amount of inert gas needed to protect a room is typically about half the volume of the room.
Overpressure flaps are used to limit overpressure by allowing the corresponding air volume to be displaced outside the room. Depending on the pressure resistance of the room, the design criterion for the overpressure flaps is typically 1- 3 mbar (millibar) for normal building construction.
The figure above shows the test setup with 1 TB SATA HDDs from four manufacturers commonly used in data centres in 2009 (3.5" drives in the enterprise storage category, 1TB capacity, 24/7 operation). To monitor the effects during the test, Siemens set the HDDs in constant operation and recorded typical performance parameters, such as SMART information, data transfer performance, and linear and random seek performance. SMART is short for self-monitoring, analysis and reporting technology; a monitoring system for computer hard disk drives to detect and report on various indicators of reliability, in the hope of anticipating failures.
During the tests, the pressure was increased via a nozzle opened by a solenoid valve. The pressure was then monitored by two pressure sensors, one for absolute pressure, and one for sensitive differential pressure.
Siemens performed a series of tests with stepwise increases in the maximum overpressure and the gradient, up to a maximum overpressure of 170 mbar and a maximum gradient of 30 mbar/s.
The conclusion was that no negative effect on performance or damage to the hard disk drives was found. As a further test, the flow-limiting nozzle was removed. Even at 220 mbar for 3 seconds with a gradient up to 100 mbar/s, no negative effects on performance or damage were found.
The test proved that no damage is caused to HDDs by overpressure created by a state-of-the-art dry extinguishing system with overpressure flaps, or by the pressure gradient in such a setup.
Is the damage caused by high noise levels?
Extinguishing systems have two major sources of noise: the acoustical alarm devices used to warn people to leave the area before the agent is released, and the discharge of the extinguishing agent itself. In the second test Siemens decided to measure the effects on the HDDs when exposed to high noise levels.
According to standards and regulations, alarming devices for dry extinguishing systems must generate noise levels between 90 and 120 dB (decibels) eg EN 12094-12:2003 Fixed firefighting systems – components for gas extinguishing systems – Part 12: Requirements and test methods for pneumatic alarm devices. Electric alarm sounders are typically at the lower end of the range, and pneumatic ones at the upper end.
During dry extinguishing system discharges, high noise levels are produced when the agent flows through the nozzle into the protected area. For some applications the noise level can exceed 120 dB. This is true for inert gases as well as for chemical agents, and depends on various design factors as well.
The figure above shows the test setup used to evaluate the noise sensitivity of the HDDs. The HDDs were placed and used in the same way as for the overpressure tests, and performance was measured in the same way.
Placed at a distance of one meter from the HDDs, a sound generator with a loudspeaker created pink noise (broadband 500 Hz to 10kHz) and 1/3 octaves (353 Hz to 10 kHz). The noise levels were then measured by a sound level meter located close to the HDD.
The advantage of using a sound generator instead of a dry extinguishing system was the high controllability and reproducibility of the test conditions. Although sound generated by a dry extinguishing system shows the characteristics of white noise, pink noise was chosen for the tests so as to avoid damage to the loudspeaker from the high-frequency spectrum of white noise.
Since the HDDs were directly exposed to the noise source without being mounted in a rack or in a computer, the tests were conducted in the worst possible conditions for noise impact.
The figure above shows at what sound level the performance of the HDDs were reduced (yellow). The test results showed that excessive noise, at a noise levels harmful to human health, could have negative effects on HDD performance that temporarily may cause periods where the HDDs are out of order (red). This level typically starts at 120 dB (at third octave level), but may for certain HDDs types and frequencies begin below 110 dB.
Most of the HDDs were found to be sensitive to noise in the frequency range of 500 Hz to 5 kHz. As expected, some resonating frequencies were found to have an even stronger impact (not shown here).
Are the laboratory results transferable to actual extinguishing systems?
In order to observe the effects when exposed to full-scale extinguishing discharges, Siemens performed further tests using the same performance measurement setup and HDDs during actual extinguishing discharges.
A series of tests were conducted with the following Siemens extinguishing solutions:
- Sinorix 1230 (agent: FK-5-1-12 / 42-bar cylinder pressure)
- Sinorix 227 (agent: HFC227ea / 42-bar and 25-bar cylinder pressure)
- Sinorix N2 (agent: Nitrogen IG100 / 300-bar cylinder pressure)
- Sinorix CDT – Constant Discharge Technology (agent: Nitrogen IG100 / 300-bar cylinder pressure / regulated valve)
As expected, some negative effects on HDD performance were observed - however, no loss of information or total destruction on the HDDs was found.
Sinorix silent extinguishing technology
After concluding that high noise levels may cause damage to hard disk drives, Siemens initiated the development of a silent extinguishing technology. The goal of the development process was to create an extinguishing system operating at the same performance level as a conventional system, but with a significantly lower noise level that would not damage hard disk drives.
In a predefined reference room, calculations were made based on a targeted design concentration of the extinguishing agent in a given flooding time. The reference room had a volume of 5×5×4 meters (25 m² floor space, 100 m³ volume) with an average reverberation time of 0.6 seconds in the frequency spectrum between 500Hz to 5 kHz. The extinguishing agent, nitrogen, was released into the room at a peak mass flow of 1.6 kg/s.
The following points present the total noise level reduction potential calculated in our reference room. When all considerations were taken into account, the noise reduction potential was 41 dB. This is a significant reduction in noise level, which confirmed the success of the development process.
The silent nozzle: using Sinorix Silent Nozzle instead of a conventional nozzle typically reduced the emitted noise by approximately 20 dB in the relevant spectrum – the equivalent to a reduction of noise energy by one-hundredth, a significant reduction.
Regulated system: using a regulated extinguishing system (Sinorix CDT) instead of an unregulated system also lowered the noise level. Our tests showed that an equalised mass flow over the discharge time led to a noise reduction of approximately 8 dB.
Acoustic improvements in the room: each room has its own acoustic fingerprint, which means the noise absorption characteristics are dependent on the frequency spectrum. The reverberation time RT60 is the time needed to deplete the sound level by 60 dB, which is one millionth of the original level. A typical reverberation time for rooms with bare concrete or glass surfaces is 2.0 seconds; for offices it is around 0.5 seconds. When the reverberation time is cut by half, the noise level is lowered by 3 dB. Therefore, lowering the reverberation time of a room from 2.0 to 0.6 seconds would reduce the noise level by 5 dB.
Extended discharge time: depending on local codes and practices, doubling the discharge time from 60 to 120 seconds can be appropriate for fire scenarios in data centres. The expected fire scenario in a data centre is more likely to be a smouldering fire arising from the electrical equipment than a rapidly developing open fire. When the discharge time was doubled from 60 seconds to 120 seconds, the noise level was lowered by 3 dB.
Correct flow direction: a nozzle pointing directly to the equipment enclosure (e.g. a storage system cabinet) might activate resonating frequencies and hence direct additional noise toward the equipment. When the nozzle was carefully positioned in relation to the sensitive equipment, the noise level was lowered by 5 dB.
Sinorix Silent Nozzle
With the Sinorix Silent Nozzle, the extinguishing process is designed in such a way that the noise level remains below the level found in tests to pose a risk to hard disk drives. Using Sinorix Silent Nozzle instead of a conventional nozzle typically reduces the emitted noise by 20 dB in the relevant spectrum.
The nozzle has a linear design that regulates the inert gas inflow via rows of tiny holes. This allows efficient distribution of the agent into the protected area, and at the same time provides the same efficiency as a conventional system. A patent has been filed for this linear design. In addition, the discharge of the agent, and hence the sound, is focused in a predefined direction, preventing additional sound-generated problems.
The Sinorix Silent Nozzle is especially developed for use with Sinorix CDT (Constant Discharge Technology ) using nitrogen and argon as extinguishing agents. Already present in our ambient air, nitrogen and argon present no environmental hazards. In addition, both agents offer excellent extinguishing properties. Using pure gases instead of a mixture of different gases also facilitates refilling because they are widely available.
CDT technology consists of a cylinder valve with a pressure-regulating function that discharges the gas into the flooding zone at constant mass flow throughout the entire flooding time. This eliminates the peak at the beginning of the discharge and thus lowers the maximum noise level.