Slashdot Mirror
ASCI White Detonates The First E-Bomb
totallygeek writes "Redefining the term vaporware, research scientists at Lost Alamos and Lawrence Livermore Labs detonated two computer simulations. ASCI White, the world's fastest supercomputer, ran the simulations of nuclear explosions. Scientists can now study nuclear weapon replacement components without violating the nuclear test ban, in effect since 1992.
Each simulation used more than 6.6 million CPU hours, which would take home machines 1000 years to complete. The data for each experiment was equivalent to 35 times the information available in the Library of Congress. ASCI White currently operates at 12 teraflops, but by early next year, Los Alamos expects to operate at 30 teraflops.
The seven month research project ended last Friday, and now the system is ready for use, after its sucessful testing."
I think the original subject for this post was a bit misleading. This was the first 3d simulation of a nuclear explosion. There have been many previous simulations of nuclear explosions, only they were limited to 2d plots of data. Nuclear explosion and fallout simulation has been the major purpose of supercomputing at Livermore and Los Alamos for decades.
Consistency please HERE
Imperium et libertas
Autocracy and freedom
Congress did not ratify the nuclear test ban treaty.
The only way i can explain this is that some people actually want other countries to develop nuclear capabilities. Which is not that far fetched actually.
It's pretty easy to find out what the computing power of Seti@Home is, just check the totals to find that in the last 24 hours, on average, the computer was running at 96.79 teraflops. Only 8x that of ASCI White.
"If I can see farther it is because I am surrounded by dwarves." -- Murray Gell-Mann
Unlike SETI@home, simulating an explosion is NOT suited for massive parallelism. The equations are partial differential equations in space and time. Very roughly speaking, to compute the speed, density, etc. of the exploding gas ball at a later time requires you to use information of the speed, density, etc. at an earlier time and at ALL spaces. You can distribute the spatial data over, say, 1024 processors, but anything more than that, the communication costs increase and you get speed degrading. In short, the nature of the equations governing the explosion imposes a restriction on the number of parallel processors you can use.
Only if you take them both as absolutes, and believe that countries can't sign onto any international law treaties without giving up all of their sovereignty.
Does agreeing to abide by state and federal law mean that individuals give up all their individual rights and freedoms? No, of course not.
Same thing.
When I was working in a Bacterial Genomics lab, I used to crave faster, more powerful computers to crunch through genomic data. This type of computing power is a dream for bioinformaticists who want to, for example, create targeted cures for bacterial disease based on specific genetic idioms.
What is unfortunate is that we have an expensive, tax-payer funded processor farm that is dedicated to the useless pursuit of studying weapons of mass destruction. A great text about the myths of US nuclear policy can be found in Michio Kaku's (with Dan Axelrod) To Win a Nuclear War. It's in the style of a book like "The Hacker Crackdown", well researched, and really interesting.
If you are interested in stopping Nuclear Weapons Research in the US, another great site is that of Nobel Peace Prize Winning group Intl. Physicians for the Prevention of Nuclear War (IPPNW). I think it's telling to compare IPPNW's site to the Defense Department's Moronic Weapons of Mass Destruction Civil Support Team web site!
I just wanted to clarify something for people thinking 'But isn't ASCII White a bunch of machines?'
Yes, It is. But they are tightly intercoupled with an IBM SP Switch that has something like 300MB (Yes, Mega BYTE) second non-blocking throughput to handle the internode communication, both at the rack (16 machine) and cluster (In ASCII White's case, it's 128 racks I believe, 128 racks of 16 4-way Power3 SPs, I've been in the same room with it but didn't touch it/work on it/have anything to do with it except go 'whoa' when someone pointed it out to me) I'm probably wrong on the interconnect speed, I think it's much faster now. I'm a bit behind on IBM's SP stuff. Spend to much time watching Myrinet.
I'd like to take a gander at the parallel coding that was done to get this kind of simulation. This can't be a batch mode program (like distributed.net and seti) like you said. It'd be quite facinating, though I'm sure they'd shoot you after you read it for that Top Secret stuff.
As a rock-in-roll Physicist once said, No matter where you go, there you are.
That is not true. Early nuclear weapons designs had severe safety problems by modern standards. It took many years of engineering and testing to solve the problems. A Hiroshima type bomb can be made to go critical by immersing it in water. A Nagasaki type bomb can explode with a measurable nuclear yield if the high-explosive lens assembly is detonated by fire or shockwave.
Mea navis aericumbens anguillis abundat
The W87 Warhead in current is already optimal! The LEP is proof of that.
Surety and command and control systems are more important in 2002 than physics simulation of actual detonations. Really, read on...
Apart from adding LAX-112 (Los Alamos stable explosive) instead of 1980s PBX-9502 and LX-17 from Livermore. (LX-l7 Is most popular in this suitcase-containable bomb), few things other than antispoof sensor and newer anti-tamper hardware need to be added.
The real problem is installing newer technologies when retrofitting to make it harder for hackers and engineers to set one off without its failsafes.
Failsafes (surety) to prevent unauthorized detotation are the most complex part of bomb engineering... not silly simulations of micro-bombs on ASCII white clusters.
These failsafe technologies have many buzzwords but are commonly called Permissive Action Links (PALs), and surety.
Many 12 digit Category F combinations currently exist but have self shutdown if too many errant codes sent to a w87 warhead.
The weapons are very good at destroying themselves electonically and physically without spilling plutonium as well.
Other than PALs, the amazing W87 also uses the Enhanced Nuclear Detonation Safety system (ENDS) developed by Sandia National Laboratories.
ENDS has lots of tamper resistence and failsafe redundancy and isolation of circuits.
A valid AMAC can be created to arm a W87 if opening the case is too challenging to a team of highly motivated engineers.
Even with codes though setting off a W87 warehead requires that each sensor is defeated.
This would include gyros, magnetometers, accelerometers, vibration sensors, and pressure transducers. Trajectory is estimated too! To prevent obvious tampering many of these were made microminiature by Sandia in EFI. (Enhanced Fidelity
Instrumentation) to replace existing larger psuedo-JTAs (Joint Test Assemblies). Lawrence Livermore National Laboratory (LLNL) current "good stuff" derives from FTU-12 (flight test unit 12) from Sandias Telemetry group (8416) and Sandias Weapons Project Group, Ed Talbot (lead), Cheryl Lari, and John Liebenberg (all in 2266).
More modern anti tamper stuff includes on-board millimeterwave radar no doubt (but classified).
Such radar would be part of the Environmental Sensing Devices (ESDs) (which determine the correct environmental conditions detonation)
The W87 weapon needs to sense spinning along its axis and pressure drop and many special measurements to detonate... even if the codes are cracked or force-probed. Sandia's MDL center 1300 (Microelectronics Development Laboratory) is tasked the challenge of defeating hardware probing and electro induced hacking of surity.
The W87 lacks coherent Lightning Arrestor Connectors (LACs), but does not need them intramodule because of its design. Multiple attempts using high energy to set of a warhead would probably just zap the circuits of the RBP (Reentry Body Programmers) or the WP (Warhead Programmers)
So at this point the W87 ( mankinds greatest technological weapon achievement) is a safe very high yield, ultra compact, tamper proof nuke.
The NSA in 1998 made it more "secure" by merely making components of its arming overly "obscure". Security through obscurity is asinine no doubt, but nevertheless the methods outlined in DIST (Defense Integrated Support Tools database) are now "top secret" even though an open crypto safe protocol would have been sufficient. The GAO, op cit., 13 August 1997, p. 8 was the last public analysis of the DIST containing the arming parts along the command chain to a W87.
You cant simulate anti-hacking anti-ICE anti-emulator anti-virtualmachine anti-clockvariance and anti-forcelatching with a damned ASCII white propaganda simulation of a mini-detonation. You need IQ and paper and pen.
Bunker busters are also lower yield than city busters, but that's because there are times you want to make a 100-ton or 1 kiloton hole in the ground without having to haul in a kiloton of high explosive or making a 20kiloton Hiroshima-sized hole in the ground and wiping out the city. Similarly, "Tactical nuke" is defined as "Designed for use in Germany" -- some of the nuclear cannon shells are designed for taking out Russian tank forces without wasting the country.
But yes, both of these are relatively scary, in that they lower the threshold for nuclear use to some thinkable, as opposed to Mutually Assured Destruction. This did deter the Russians, but it also made it easier for the US to step on Russian satellites so it wasn't decreasing the chance of war, just changing the terms and the probable battlegrounds.
Bill Stewart
New Fast-Compression-only CPR http://preview.tinyurl.com/dy575ks
It's "Moore's Law", and it does not state that CPU speed doubles every two years. Moore's Law states that transistor density doubles roughly every 18 months.