Slashdot Mirror
Coolest Cluster Ever
sw155kn1f3 writes "Scientists at Los Alamos National Laboratory built a cheap (less than $1k per unit) 294-unit Beowulf claster dedicated to run astrophysics calculations. According to their website it's 85th fastest computer in the world. Seems cool and promising as it made with cheap components and off the shelf hardware."
Imagine a single unit of these...
Imagine a Beowulf cluster of these...Beo...shit - I knew this joke would have to end at some point.
Cue The Sun...
They didn't even use a rack mount solution, they used regular Shuttle XPC SS51G Mini-PCs
;)
I thought Shuttles Mini-PCs were cool before but this really resets the scale... Now where is the HOWTO for this thing?
.: Max Romantschuk
I don't really think you can really quantify coolness in general, but I fail to see how the fact that this thing is cheap makes it all that cool.
Perhaps if it was going to run simulations of ultra-low temperature physics... get it? haha. I kill me.
autopr0n is like, down and stuff.
It would be possible for a group of people, not necessarily a small group, but not necessarily huge, either to repeat this. 100 people, each with $3,000, could do it. The group would need to find some space to house the thing, and would probably have to do it in a climate where it could be relatively naturally cooled, which definitely rules out Phoenix. The computer would then be one of the fastest machines in the world.
Granted, I don't know what the hell they'd do with the computer, but it would be kind of cool to be on the list.
IBM had PL/1, with syntax worse than JOSS,
And everywhere the language went, it was a total loss...
and a "heat pipe" instead of a fan!
Actually, the heat pipe doesn't replace the fan, it just lowers the number of fans used in the system, since the case and processor fan can be combined.
Tom's Hardware has a review of one of these things (not the same model though)... have a look.
.: Max Romantschuk
Take a look at this room A 1000-Pentium Beowulf-Style Cluster Computer half way down the page.
Analytic & algebraic topology of locally Euclidean meterization of infinitely differentiable Riemmanian manifold
2. ???
3. SLASHDOT!
Laugh, it's a joke.
They could spend their $3k on something actually needed by 100 people thinking about spending $3k to share a spot on a fastest computers list.
Like a prosititute.
You go to your high school reunion, what's, more impressive, the "Hugh Hefner" 100 $3k prostitutes that come with you, or the "Bill Gates" story about the 300 1k computers in your mom's basement?
paintball
The photo alone is worth surfing over to the article. As Socrates once said, "what a rack!"
:-)
But now that they've got the 85th fastest computer, what will they have to do to maintain that coveted position? I imagine the people who are running 86th are rushing out to buy more nodes. My own computer is the world's 27,385,422nd fastest, and I'm battling like crazy to get to 27,385,421.
I thought satire was supposed to be funny.
I sincerely hope that soon small-to-medium enterprises can own supercomputers. With all the low budget physics stuff going on at Universities around the world, cheap supercomputing can only be a good thing.
Actually they can with software like that from Dauger Research, Project Appleseed and Wolfram Research with gridMathematica
The cool thing here is that this code can be run on all of the desktop computers that already occupy companies and universities world wide allowing for easy access to supercomputer level computational speed (for those problems that can be attacked using parallel computation of course) using the same computers normally used for productivity.
Very cool.
Visit Jonesblog and say hello.
I really like the picture on the sites frontpage.
:)
:)
:)
.au dollars for all our procs (and maybe switch - I don't know the details) - half the performance, half the cost.
i can imagine the small size of the Shuttles being an advantage, not to mention the "coolness" factor looking at it. (i assume the "cool" in the intro refers to emotion and not teperature!!)
But getting computation done cheap is no longer the challenge. It's getting the data from one node to the other. They still need "custom" expensive equipment for this.
I see they use 3com gigabit ethernet. having this 300+ gigabit switch capability is not "cheap".
Until one can buy this kind of networking equipment for really cheap, we shouldn't mention things like "of the shelve Beowulf super computer in the top 100".
Up until 3 hours ago, we said we were the only machine in the top500 dedicated solely to astrophysics. Now we are one of 2
But we use 180 processors, @ 2.15 GHz, and get about 0.90 Gflops/processor, whereas they get about 0.88 gflops/proc, getting us in the top 180th.
The difference being they have faster memory, and we have a big badass switch. They have two switches, with something like only 10 gigabit between the switches! We have 250gbit within our one switch. a third of our nodes have 2 gigs RAM, and we also have room for upgrade to more nodes on our switch, and they don't. So, in the words of Nelson "Hee haw!"
When they say $1000 per proc, they are not factoriing in their two switches. This will bring the price up to about $500,000, unless someone is donating a switch or 2
We have about the same cost ratio - something like 250,000
What's the bandwidth of that trunk? Also, what's the capaity of the connections between each 16-port card and the backplane?
Just curious... suppose all the units on a 16-port card have 1Gbps each, but only 8Gb total to the backplane. Then the backplane, in turn, has only 8Gb to the other switch. These are just made up numbers, but how would beowulf handle it? Can it group jobs requiring higher communication throughput onto the nodes which are closer to eah other? Does it have to be told the topology, or does it figure it out?
Sounds like they have 10 ports @ 1Gbit each free, so get about 10Gbit between the switches. When we (position 180 on the latest top500) were investigating thin-tree connections, we thought that we might be able to effectively run one job on a third of the nodes, another on the second third, and miscellaneous jobs on the 3rd (since we have two or 3 researchers who like to use lots of resources, and people like me who only need a single proc at a time). So you just partition the nodes and only allow your mpi jobs to sit on one group at a time.
Then there are things like openmosix which deal with topology automatically somehow. They will try to calculate the speed of the interconnect and the various nodes' procs (if heterogenous), and work out the distribution in the most efficient way. I am going to try to convince my sysadmin to try out openmosix on the lesser-used nodes of the cluster, because there is a feature of it that I think one of us might like - the combination of memory of the different nodes in one big contigous space, but right now we are busy cleaning up after the upgrade.