Slashdot Mirror
Army Contractor To Build A 1566 Xserve Cluster
olePigeon (Wik) writes "MacCentral has an interesting article on a new computer cluster. From the article: 'Apple Computer Inc. will announce on Monday the sale of 1566 dual processor 1U rack-mount 64-bit Xserve G5 servers to COLSA Corp., which will be used to build what is expected to be one of the fastest supercomputers in the world. The US$5.8 million cluster will be used to model the complex aero-thermodynamics of hypersonic flight for the U.S. Army.'" alset_tech was one of the many readers to point to
CNET's version of the story.
for America's Army!
well, you know, they saw this spaceshipone on tv tonight and thought that hey, "we want one of those too".
or possibly "wtf how does that thing fly??".
world was created 5 seconds before this post as it is.
"US$5.8 million"
I'm a grunt in the USMC (former computer geek...who would have figured?)
Anyways... I'm about to go *back* to Iraq in September.
The high brass has some f*ed up priorities some times.... the army has $5.8mil to contract out *research* to some company for technology what.... 10-15 years away at the minimum?
Meanwhile the Marine Corps is scraping nickles and dimes to get us basic equipment the army has had for most of a decade.
Hell, when we go to the field to train, we often have to yell "bang! bang!" because we don't get enough (or any) blank rounds for training.
Imagine if they took just ONE Osprey off the project..... maybe then I wouldn't have a hand-me-down-from-the-army m16a2 (does the army use them anymore?)
Army Contractor To Build A 1566 Xserve Cluster You totally misspelled "1337."
"Physics computations and weapons simulations so good looking, you're going to want to lick them." - Steve Jobs
"Uh, we'd advise against that sir." - Army colonel
"But he SAID I could lick them! Ooh, red, yellow and green WMD icons!" - G. W. Bush
(Translation:
This is...
O
U
T
R
A
G
E
O
U
S
!
Not a lot of bombs that fly hypersonic.
Tank and artillery shells, on the other hand...
Why yes, I AM a rocket scientist!
...that the Army is buying.
Tubal-Cain smokes the white owl.
I think that the reason you don't have any good equipment is because the USMC and the USN blew their allowance on a wasted IT upgrade from EDS. You know, that 7 _BILLION_ dollar contract which has already triggered SEC investigations for fraud.
Want to bitch about not having bullets? Look to your own leadership and stop whining about how the Army is going to spend its budget.
" Isn't hypersonic flight research better suited to the Air Force?
How about some hypersonic sub-orbital artillery with your fries, Sir? Granted that's the navy version, but whatcha wanna bet that the Army could put a land based platform to good use?
You need a FREE iPod Nano
The A-10 is not operated by the Army. The Army tried to buy all the A-10's from the Air Force, but the Air Force didn't want the army to operate them. So they gave them to the Air Force Reserves.
It all comes from the WW2 era pissing contest which made the Air Force a separate branch from the Army. It is a pretty silly distinction, to everybody except the Air Force, to whom it is Holy Writ.
And you're right re: the bombs. That was my original (oblique) point. : )
Why yes, I AM a rocket scientist!
For render farms and low interconnect requirements, blades are really popular because of manageablity and density (though I am curious about the manageability of Xserves beyond OS management, i.e. service processor presence/capabilities).
As to the comment about no 'mac' blades, it is true, but if you are a big fan of power architecture, IBM has announced JS-20, a power based blade, which has the 970 (same as G5), but only at 1.6 GHz (ironically enough, IBM doesn't seem to sell anything at the clock speeds Apple gets to sell at, and they are all IBM's chips...).
The cost of blade solutions with myrinet or infiniband solutions is significant. Otherwise, most chassis' I see communicate externally through an oversubscribed ethernet switch. Ethernet is inherently sub-optimal, but oversubscribed ethernet is particularly troublesome for some of the fine-grained parallel applications (embarrasingly parrallel applications, of course, don't care, and rendering is one such application).
Add to this a lack of expansion capability (i.e. IBM blades can take one daughterboard, so there is not any possibility of, say, having a fibre channel *and* myrinet adapter in a blade server.
The only thing I'm aware of with respect to high-performance interconnect solution for blade servers available today is to get IBM blades with Myrinet daughter boards and an optical passthrough module. Ultimately, it can really reduce cabling for things like ethernet, kvm, etc etc, but those myrinet cables are still going to be a tad unwieldy (80+ wires to the cabinet, even if they are fiber cables).
I actually want to see a solution that would aggregate, say, 1X infiniband to each blade into 4 4X connectors, no oversubscription and much sturdier and fewer cables.
XML is like violence. If it doesn't solve the problem, use more.
The Army is indeed testing a hypersonic anti-tank missile that launches from a box mounted atop a HMMWV. A buddy of mine worked on the guidance system.
As a matter of fact, a lot of Lockheed Martin's next-gen missiles are kinetic kill vehicles: No explosives, just a lot a lot a lot of velocity.
Why yes, I AM a rocket scientist!
One thing to note is that vector processing doesn't help the G5 on the linpack benchmark (what the top 500 list uses). Altivec only helps single precision calculations. Good for games and graphics, but not desirable for many scientific calculations.
It's the fact that the G5 can dispatch two floating point operations per cycle (like the Athlon's fpu) and that it has a fused multiply/add instruction that can be done in 1 cycle. This effectively gives it the ability to do 4 flops/cycle.
So the theoretical peak is given by 1566 xserves * 2 cpus each * 2 GHz * 4 flops/cycle = 25.056 teraflops/s
And the reason for this is because the mass is traveling faster than the wavefront of the explosion would. Seriously an explosive would only disrupt the kinetics. There was a call for shoulder fired hyper-velocity missiles a few years back, I have no idea what happened with them.
Bad Panda! No Bamboo for you! In matters of importance ACs will not be responded to. Want to say something critical,OK