Slashdot Mirror
Sun Donation Spurs Linux Cluster at Purdue
An anonymous reader writes "Purdue University, with a $3.6 million gift from Sun Microsystems, is giving recycled PCs new life as a computer cluster that makes high-performance computing power available in undergraduate classes. 'Previously, my students could only do what I'd describe as 'proof' animations - small, low-resolution and not presentation quality,' [Professor Richard] Paul said. 'With access to this computing power, the students will be able to ship their software files of instructions to the Linux cluster, and it will come back in three or four hours with modeling, lighting and animation. Students will get to experience the whole thing in terms of scale and presence, and they can do longer animations.' More images of the current Linux cluster and other servers at Purdue are out there."
Sun 'Origin' High-Performance Servers and Supercomputers.
Your hair look like poop, Bob! - Wanker.
They're SGI Onyx 2 computers, something like an Octane with a lot more upgradability. you wouldn't be able to fit multiple Infinite Reality 2 pipelines in an Octane... You can find them here, though I wouldn't want to pay the power bill for one:
http://www.sgi.com/products/remarketed/onyx2/
GPL: Free as in will
I have a good friend who is a CG major at Purdue. There were always difficulties in trying to get even simple animation projects rendered in a timely manner. This is a great resource for students who are really trying to get their projects done in a snap, who can see the results of their efforts a little sooner than a full day later.
They're all identical because they're the old computers from the ITaP Computer labs. As a Purdue Student (McCutcheon North REPRESENT!) I can't say I'm too broken up about seeing them replaced with the newer (and faster!) Dells.
That 72% of all statistics are madeup on the spot?
"These machines from sun suck down the electricity and provide measly amounts of gflops as thoer benefit"
Umm numbers please?
"Each month the elctricity bill could have bought them 4 more dual g5 macs."
Again where's your proof?
How about next time more facts and less fanboi.
If you wanna get rich, you know that payback is a bitch
Sun never bothered to port their UltraSparc beowulf-like clustering system to X86, and they stopped ripping off Linux code after the whole ethernet module fiasco a few years ago.
Hence, no X86 clustering support with Solaris.
You can't judge a book by the way it wears its hair.
There are companies that provide render resources. They are great if you are a small studio and only go to final render once ever so often. They are expensive if they are your only resourse.
For a past project, we priced out outsourcing our final renders and discovered that it would have cost us $6M to render on $1M worth of hardware for about 4 months. Prices may have changed since. We dropped the money and brought in 400 boxes because we could have reused the machines for future projects.
But studios of 10-20 people, its not a bad idea. When it hits crunch time, its always good to have a couple hundred extra machines to get stuff done.
-Tim
-I just work here... how am I supposed to know?
You can use Maya 5 or 3dsmax 5. You submit the jobs via specific lab computers and get email notification when the job begins and finishes rendering.
More details are available at pete.purdue.edu.
These ITAP (formerly PUCC) clusters are in the Math buidling. You're thinking of some of the ECE clusers in the MSEE machine room and in MSEE214. We still have those too, but they're limited to ECE research use (i.e., me!).
The guy who ran the cluster in Civil (Moffett, quoted in the article) moved over to ITAP and helped start this thing about 1.5 years ago or so.
As far as I know, the donation from Sun had nothing to do with creating this cluster--that's just media spin. It may have something to do with growing it or new applications, but the cluster has been around for a while. It was created to help offload work from the IBM SP, which had a multi-week wait for CPU time until they brought the recycled machines online.
The idea is pretty solid though. Purdue gets rid of computer lab machines every 2-3 years, and rather than sending them off to salvage (literally a dump), they stick them in a machine room and use Condor/PBS to schedule jobs. Very useful to those of us grad students who need a lot of CPU throughput to do work. And believe me, a lot more than just rendering projects is going on--folks from many departments use these things.
And since the university has it's own power plant, it doesn't have to pay retail prices for the elctricity that runs them (or anything else).
For those who are stat-hungry, this system currently has about 700 total CPUs on the main system, all Pentium IIs and Pentium IIIs. There's also 200 Athlons, but they're technically owned by somebody else (i.e., they're not recycled). As of right now, the 1-minute load on the main system 893 (we're pretty busy right now). There are 598 jobs awaiting CPUs. I've personally run 250 jobs today, and I'll be kicking of another 200 in about an hour (big research deadline soon).
There are several hundred more machines available for use, but they are not up yet due to lack of floor space, cooling capacity, network connections, and/or power. Much to the annoyance of those that could really use more compute time.
The PCs themselves came from university computer labs, which is why they had ZIP drives in them. Purdue has ZIP drives in all lab machines, and the article says these machines were used for 2-3 years in the labs. Apparently the Sun donation was for the server hardware, and not the actual cluster machines.
What about power wasting ?
this got marked -1 off topic, but power consumption is a fixed cost associated with running a cluster and VERY relevant in budgeting a cluster.
Electrical costs make this a White elephant gift!
The Dual G5 VT cluster (1,100 dual g5 macs) is not only rated as the 3rd fastest super computer on at www.top500.org next november, it is also one of the cheapest per kilowatt hour to run, not super cheap, but cheap enough.
These machines from sun suck down the electricity and provide measly amounts of gflops as thoer benefit.
This department would have been better off getting a handlful of g5 macs.
They can do 16 GFlops peak if calculating FMADD pairs.
Why?
Because the g5 has TWO FPUs per chip and heir are two chips.
And each fpu can do a combined multiply-add per cycle and there are 2 billion cycles per chip and two chips in the 2999 dollar macs.
plus you can wedge 8 GB physical fast ram in them, and they come with fast dvd burners to burn and rip porn movies.
16 GFlops on a modern mac that chews up not too much electricity for cpus alone, make this gift from sun a white elephant.
Thats not even counting the 128 bit vector processors on the g5 (Altivec). Those things offer SIMD using over 110 different amazing opcodes each.
Each month the elctricity bill could have bought them 4 more dual g5 macs.
I think you got it the wrong way. AM radio is the one with the long range. FM radio has a shorter range, because it takes more power and can only travel in straight lines. AM radio bounces, and can therefore travel beyond the horizon.
For this reason AM radio is used on sea when satelitte-transmissions are too expensive.