Home page logo
/

nanog logo nanog mailing list archives

Re: Points of Failure (was Re: National infrastructure asset) (fwd)
From: George William Herbert <gherbert () retro com>
Date: Tue, 25 Sep 2001 16:05:08 -0700



Some of this is drifting off topic.  I have also juggled the order of
some of David's points in order to address them in order of most-on-topic
to least-on-topic for this list.

David committed to electrons:
Unnamed Administration sources reported that George William Herbert said:
There are many many angles to all of this.

And who will pay?

In the end the customers will, of course.

For any specific protective measure, there's a balance between
risk and cost to protect against it, and for all risks there's
a total balance of how much the customer is willing to pay
overall period.  We now know that the risk spectrum is both
different and wider than it was.  That now has to be assessed
both technically and financially (cost of protecting, costs and
risks of sustaining losses, costs customers are willing to bear)
by all of us in the business now.  The answers will not be
the same for each business.  The answer probably will be more
push towards more hardening the more central to the net
infrastructure a particular facility or business is, though,
as a sweeping generalization.  We've now come very close to
losing major exchange points or telco centers of equivalent
criticality several times.  The consequences need to be addressed.

The answer is obviously not going to be every colocation facility
all the way out to your POP in Elko, NV being armored enough
to withstand a nearby nuclear detonation.  But how we do things
structurally probably needs to shift a lot as a result.
Too many people in the industries have blown off physical risk
too much for too long.

Not that I'm recommending every telco facility be hardened to
resist a direct jumbo jet hit.  But it's not that ridiculous
a task, and lesser hardening (half or 1/3 that level) would be
downright reasonable compared to the other costs of making these
sorts of buildings for these roles.  A 2 foot thick reinforced
concrete wall, for example, might well be extremely reasonable.

But every n of wall thinkness is f(n^2) less interior space;
f(n^3) more cost to build, and f(n^8) more crap from local
pols who really want retail space downtown instead of tombs.

For small n compared to building dimentions, it's not n^2 but
only n additional ground footprint.  Ground footprint of my
theoretical building without armor is 200x200=40,000 ft^2;
with it, it's 220x220=48,000 ft^2.  20% more total. 

Thicker reinforced concrete walls are actually cheaper on
a per pound basis than thinner ones are, believe it or not.
Forms don't care how much space is between them, and form
prep work is a lot of the cost.  The actual rebar inside is
porportional to volume/mass, as is the concrete cost, 
roughly (they both get marginally cheaper as you get to
larger jobs and scale improves).

If the local pols care, then you can either not build your
data centers downtown, or put darkened glass windows on the
outside of your concrete building to make it look like offices
(this is actually reasonably good disguise and has been
used in actual hardened facilities before).  Another option
is building-within-a-building; have offices around the 
perimiter, with the datacenter behind an internal solid
concrete wall with few openings into the datacenter.

Let's look at the "worst credible case" though, jet aircraft.
Nuclear reactor domes are rated to survive jumbo jet hits.

Better check this one out; there is considerable dispute
of late on this very point.. A F4 !=747 after all.

The F-4 was used for one particular demonstration and test
(the famous jet-on-rocket-sled-into-concrete-block video clip).
The design criterion was a moderately loaded 747.

We now diverge into well and truly off topic analysis...

Looking at the problem, you have a 20x200 foot fuselage weighing
about 500,000 lbs with fuel and baggage, a (4-0.5)x200 foot
wing weighing 200,000 lbs with fuel, and 4 8 foot diameter
engines which weigh about 12,000 lbs each with fairing and
mount hardware.

The only parts that are really "hard" are the engine shafts
and the main landing gear.  The rest is pretty weak.

The concrete wall mass density is around 2.3 tons/cu meter,
or around 7 tons per square meter of outer wall.  That works
out to 1,400 lbs/ft^2.  The jetliner body at 500,000 lbs and
about 315 square feet has a mass loading of about 1,590 lbs/
ft^2.  That will crumple in over an airframe crumple distance
of 200 feet, roughly.  The engines have a mass loading of around
240 lbs/ft^2, but will concentrate most of their mass over about
10% of their area briefly and are only about 15 feet long. 
At an airspeed of 250 m/s, (562 mph, 500 knots) the body will
impact over about 0.25 second with typical forces of around
50 million lbf, which is trying to shear a 20 foot diameter
10 foot thick reinforced concrete plug out of the larger
wall structure (to use the simplest but generally quite
accurate penetration model).  That's 630 square feet or
about 90,720 square inches, or a shear force of about 550 PSI.
Good reinforced concrete will withstand more like 2,500 PSI
minimum, and some of the fiber strand reinforced varieties
along with rebar / mesh reinforcing are several times stronger.

The center secton of the wing hitting at mid-fuselage is the
largest single impact transient in the collision event, but
even those don't penetrate.

The engine impacts are significant primarily because they are
largely focused via the engine shaft; they will be decellerating
over about 3 meters in 0.025 second, with force loading on the
wall of 12 million lbf over a roughly 8 foot diameter plug
with surface area around 250 square feet or 36,000 square in.
The shear forces on the plug walls are about 333 PSI.  With much
of the engine inertia effectively transferred forwards via
the central shaft, this can transiently be 10 times more
focused for parts of the impact, to 3330 PSI, approaching
the failure point for the wall and perhaps exceeding it for
limited areas of traditional reinforced concrete.  The vast
bulk of the engine mass should be stopped by the wall, however,
and higher strength reinforced concrete will resist it adequately
to prevent penetration.

Jet aircraft are remarkably poor penetrators, in military terminal
ballistics terms.  Ten feet of good reinforced concrete, at
$400/cu yd or so installed, will do the job nicely, though there
may be some spalling inside the aircraft won't penetrate.  For a 3
story, 120,000 sq ft 200x200x30 ft building that would be about
$9.3 million, about $77.70/ft^2.  

No help. You'll have intact but long since roasted into
atoms building, hardware and employees. How many KWH of heat
from the 57,285 gallons of kerosene?

Not a problem...

Most of those KWH go up in the fireball and atmosphere,
and the 40,000 ton concrete shell is a great heat sink
and insulator to absorb and resist the rest.  Much thinner
walled bunkers are used for nuclear weapons installations at
airfields, which have to survive having loaded bombers
and fighters crash on the bunker and not cook the weapons
inside, for example.

As a minor aside, the example building I posted was woefully
mis-optimized if you're worried about such hardening issues.
You want a building that more closely resembles a sphere
or cube (minimum surface area for contained space) rather
than a relatively flat slab.  Something more like 120x120x120,
5 stories up 3 below ground with 15 foot floor heights,
would be much better.


-george william herbert
gherbert () retro com


  By Date           By Thread  

Current thread:
[ Nmap | Sec Tools | Mailing Lists | Site News | About/Contact | Advertising | Privacy ]