Virginia Tech assembles Top100 Mac supercomputer cluster

Tue, 07/22/2008 - 4:28am — Jonny Evans

Buried beneath the hype surrounding iPhone and Apple's growing Mac market share, the company still offers server-grade solutions that are making some dent in the high-end research markets - and now one of the company's early champions is adding another Apple server farm to the collection.

Virginia Tech has deployed a new 29-teraflop Mac supercomputing cluster that's based on 324 Mac Pros, it has been revealed. The cluster is at the Center for High-End Computing Systems (CHECS) within the Virginia Polytechnic Institute and State University. The system would have been ranked in the Top 100 systems of the June 2008 Top500 list of supercomputers.

Ensuring the Macs are all talking to each other, Mellanox Technologies' 40Gb/s InfiniBand technology has been deployed to interconnect the machines.

Thad Omura, vice president of product marketing at Mellanox Technologies, said: "In 2003 we partnered with Virginia Tech to build the first 10Gb/s InfiniBand large-scale cluster that was ranked number three on the Top500 list at the time. We are excited to partner with Virginia Tech for the first large-scale cluster installation using 40Gb/s InfiniBand."

“Our mission is to build computing systems and environments that can efficiently and usably span the scale from department-sized machines to national-scale resources, and will meet the day-to-day needs of computational scientists,” said Srinidhi Varadarajan, Director of CHECS at Virginia Tech. “Mellanox’s 40Gb/s InfiniBand technology brings the necessary capabilities to our research activities and will be the focus for the design and deployment of the next generation of high-end systems.”

( Filed Under:

Apple Inc News

)

Comments

Lame. You could do the same

Tue, 07/22/2008 - 5:45am — LD (not verified)

Lame. You could do the same with consumer level gaming GPUs for maybe 1/10 the cost or less. Maybe 1/100 the cost. Just look at the Fastra project or, if you go high-end, the NVidia Tesla. Both are capable of around 1 teraflop. Combine 30 of those, for much less than the cost of 324 Macs or PCs, and you'd kick this cluster's ass.

Lame?

Tue, 07/22/2008 - 6:07am — 9to5Mac Noob (not verified)

People with the technology to build supercomputers don't just throw money into something without research first. There are reasons why they have used this method, perhaps it's more suitable to the application, perhaps it's easier to implement, or maybe they don't need the speed or parallel processing of GPU's, but the reliability of standard cores.

Or maybe they should have just asked you first...

Bear in mind that they could

Tue, 07/22/2008 - 6:13am — Reply (not verified)

Bear in mind that they could upgrade to Snow Leopard, and swap out the CPU to upgrade, and also add more RAM, or add any custom made GPU system into those Macs. And those computers all work as separate machines i'd imagine, so it's the price of a very nice 300 Mac Pro suite, that can also do clustered supercomputing.

Yeah, lame

Tue, 07/22/2008 - 6:15am — LD (not verified)

Read up on the Fastra project. They had been using a similar cluster for a long time, until they determined it could be done better and cheaper.

Look into the folding@home project as well.

A large part of what Virginia Tech is doing is number crunching, something massively parallel GPU computing excels at.

Admittedly, it's not for all scenarios, but most likely they could utilize it and save a few million in the process.

Wow...

Tue, 07/22/2008 - 11:10am — 9to5Mac Noob (not verified)

Freddie Mercury came back to life, and is now in Blacksburg, building clusters....

Who knew?

System X

Wed, 07/23/2008 - 6:10am — 9to5Mac Noob (not verified)

Back in 2003, VT built System X, which is an 1100 node, dual processor cluster. It was ranked the 3rd fastest in the world. It became modeled after by many others wanting super computers because of the processing power to cost ratio. I think VT knows what they are doing.
System X is currently ranked 280.
http://www.top500.org/system/7286

not to dis VT or apple but

Wed, 07/23/2008 - 1:18pm — 9to5Mac Noob (not verified)

not to dis VT or apple but IBM's new roadrunner can do a pedaflop of processes per second. so in laments terms 1,000 Trillion processes per second. Don't get me wrong I love apple as much as a fanboy but I have to hand it to IBM on this front.