Fire Block

Fire Resistant Concrete Block Backup Protection

Important Notes

Current configuration does not allow embedding of the file fb1.swf because of its mimetype application/x-shockwave-flash.: fb1.swf

Assembly of the fireblock, cables not shown

(1) Lay down first tier of concrete blocks on flat floor.

(2) 2nd tier core block, horizontal holes.

(3) Add drives and multiple layers of aluminum foil wrapping.

(4) First tier of 2 liter pop bottles, filled with water and an adhesive metal foil lid designed to burst at low pressure.

(5) Rest of second tier and third tier of concrete blocks.

(6) Second tier of 2 liter pop bottles and fourth tier of blocks.

(7) Some sort of cover and banding around circumference, not shown.

See attachments for Pov-ray source files.

Note for testing - I have high-temp-insulated thermocouples and meters for testing, as well as an accurate temperature controlled test oven. A protégé can video-record the test burns.

A Seagate FreeAgent GoFlex Desk 3TB USB3 drive is less trustworthy than other backup drives. But it draws 10 watts while moving 2GB/minute, draws 6 watts idle, and nothing when powered off. The latter might be arranged with this board for $16 (not USB! Crooks!). With two or three drives used in alternation there would ample redundancy to correct for potential flakiness.

Cooling During Normal Operation

My backups take about 4 hours to connect to many machines, five offsite, three at the wrong end of a 80kB/s DSL link. With faster links the backups would be much quicker. Initiallizing the external hard drives can take days, but that can be done before the block pile is built. Initialization is when infant mortality is most likely to kill a hard drive, so it is best to run the drives at 100% for a couple of days before building the fire block.

4 hours of 10 watts is 144 kJoules, 35 KCal, heating 5 kilograms of water by 7C, or 25 kilograms of concrete (the specific heat of concrete is 0.2 cal/g-K) by the same amount, even if dissipating that heat takes a day. As we will see below, the thermal conductivity of the concrete interconnect webs are 3W/C, so the core would heat up less than 4C if run continuously at 10W, forever.

The computer driving the backup is outside the pile, and expendable - it will dissipate a lot more power than the drives. If the computer goes during backups, the drives lose power (they are normally powered off). As an alternative, a single board computer such as a PC Engines APU 1C could be inserted into the pile, so only an ethernet cable and DC power cables connect to the outside world; however, this adds to the internal power dissipation, up to 24 watts maximum.

Construction for Fire Resistance

Fireproof safes are mostly concrete - very expensive concrete in a heat-conducting metal box with a door and a lock whose combination could get lost. A 16x8x8 cinderblock costs $1.42 at Home Depot and weighs 20kg. A stack of 24 blocks arranged in a 24"x32"x32" stack, with a couple of backup drives in the center sealed against water, would weigh 480kg. The holes in the blocks around and above the center are filled with 22 2 liter pop bottles full of water, which would boil and carry off heat in a fire. In normal circumstances, the heat capacity of the assembly is ( 480*0.2 + 44 ) = 140 kCal/C, 678 kJ/C, so it would heat less than 0.3C during a typical backup run. Of course, it would heat more in the middle and less at the edges.

When it gets hot enough, the adhesive seals on the pop bottles will burst and make a lot of steam, carrying off heat. In a fire, the pile might get drenched with firehose water. Fireblock will take up a lot of room, and it will be a chore to disassemble the stacked blocks to get at the drives. But it is cheaper and a lot more survivable than a fireproof safe. A safe has much less thermal mass and has a lock that could be damaged by heat. Bad guys with unlimited physical access might be able to disassemble the pile and get at the drives, or access the USB hub and get at the data in the drives, but that will be time consuming and more subject to discovery and intervention.

For extra fire resistance, wrap the pile in thick house insulation and another layer of concrete blocks. I imagine a fire will burn out before a lot of heat makes it into the smaller block pile shown, though. I need to learn more about the time-temperature behavior of fires so I can calculate the response.

Not perfect, but cheap and better than nothing. I have an unused fireproof safe, but I would rather use that for papers, and not drill holes in it.


This NIST report local copy about an instrumented house fire has a graph on page 13 of radiant loading rising exponentially from ignition to 8 kW/m², dropping to 6 kW/m² after 100 seconds (opaque smoke and CO₂?), then plummeting to near zero after 350 seconds and the beginning of fire suppression. The outer surface of the fire block (5 sides) is 2.8m², so in these circumstances the fire deposits (8000x100 + 6000x150)*2.8 or 4.8 MJ on the concrete and water. At 678kJ/C, the temperature rise would be 7C if distributed uniformly (most will in fact heat the outer surface, and little will penetrate to the core block).

The actual thermal penetration will be through an outer 1.5 inch thick ring of concrete, through 15 1.5 inch wide, 5 inch long webs, 32 inch tall webs of concrete, to a double inner ring of concrete effectively 3 inches thick, past the water bottles which will limit temperature at the webs to 100C until the water turns into steam. Assume that the inner ring and core block weighs about 110 kg, the connecting webs weigh 15/35ths of the 400 kg of outer blocks (170 kg), and the outer facing weighs the remaining 200 kg.

Assume concrete has a thermal conductivity of 0.8 W/m-C. Without the water, the 1.5 inch thick outer wall plus the outer half of the connecting web weighs 280 kg, and has a thermal capacity of (280k * 0.2 * 4.84 J/C) or 270KJ/C, so assuming 17 kW radiant heat incoming (and soot-blackened so complete absorption) the outer wall will heat 4C per minute, increasing from 20C to 100C in 20 minutes. The thermal conductivity of the web is (15x1.5x32/5 inches)*(0.0254 meters/inch)*0.8W/m-C or 3W/C, so if the inner core is at 20C then a 100C outer facing will feed the inner core with 240 watts. If the outer shell heats to the full fire thermocouple temperature of 600C from the NIST report, then the maximum heat flux is 1800W. If we include half the web with the inner core (200kg), with a heat capacity of 190 kJ/C, then the core would heat at 0.57C per minute, 34C per hour. The fire would have to rage for 1.5 hours at 600C before the inner core reached 70C and exceeded the operating temperature of the hard drives. There's probably not enough combustion fuel in a house to burn that long (even with my huge basement library, a book-burning fanatic's dream).

There's 44 kilograms of water built into this system. If half of that turned to steam, the total heat capacity of the water plus steam is 80C*4.84J/C-g*22 kg or 8.6MJ for the heating to 100C, and plus 2.2MJ/kg * 22 kg or 48.4MJ for heat of vaporization, for a total of 57 MJ carried off. At 17 kW incoming heat, that is another hour of survival.

All of these are worst case assumptions and crude approximations. In reality, the inner and outer shells have thermal resistance, the drives may survive higher temperatures, and the nearest fire station is 1.3 km away from my house. I suspect this setup will survive major regional disasters that no human outside of a deep bomb shelter will.


Dirvish wiki: FireBlock (last edited 2020-04-09 21:19:30 by KeithLofstrom)