Need Help Putting Computer Together?

Maybe these guys can help.:D


Computer is built in a flash
In a daylong project, Purdue teams install a 2nd 'super' device

By Eric Weddle
Posted: July 22, 2009 WEST LAFAYETTE, Ind. -- Purdue University built a supercomputer in one day -- again.

Within about five hours Tuesday, teams of volunteers pulled hundreds of servers off trucks and began combining them to become the biggest campus computer in the Big Ten.


More than 200 information technology staff helped assemble the supercomputer, named Coates, from more than 10,000 computer processors. The completed computer will allow faculty to crunch massive data sets for cancer research, climate modeling and other uses.

Coates is called a "community cluster" because 30 Purdue faculty contributed research money to fund the purchase of nodes, or servers, said John Campbell, associate vice president of the Rosen Center for Advanced Computing at Purdue.

"The faculty members' machines all operate as one big machine," Campbell said. "If one faculty is not busy, another can borrow those nodes to do computation."

Last year, a supercomputer named Steele was built in one day with 6,500 cores, and it continues to be used. The new computer was needed because more computational power was requested by faculty, Campbell said.

During Tuesday's installation, computer colleagues from Indiana and Michigan State universities, the University of Michigan and the University of Iowa visited to learn how the computer is assembled so quickly.

Coates, which is in the basement of the Mathematical Sciences Building, contains 1,280 dual quad-core servers, each containing the equivalent of approximately eight computer processors. Each server costs about $2,000.

Mathematical stats:
1280 x 8 = 10,240 total processing cores

That's a lot of processing power.

Quote:

---

Originally Posted by Reggie

Mathematical stats:
1280 x 8 = 10,240 total processing cores

That's a lot of processing power.

---

Can do do the comp in layman's terms?


How many bags of M&M's and cans of Mountain Dew?;)

Dont forget the Hot Pockets.

While I'm sure they have developed a great cluster for doing campus computing task I am at a loss as to how this can be called a supercomputer that works as a one big machine. Unless these guys have developed some super optical data bus I don't know how they can harness all or those processors in a parallel fashion to harness the computing power of all those processors. All those processors need a lot of databus bandwidth :confused:

A lot of hot pockets, that can be sure.

Quote:

---

Originally Posted by kokopup

While I'm sure they have developed a great cluster for doing campus computing task I am at a loss as to how this can be called a supercomputer that works as a one big machine. Unless these guys have developed some super optical data bus I don't know how they can harness all or those processors in a parallel fashion to harness the computing power of all those processors. All those processors need a lot of databus bandwidth :confused:

---

Clustering software can split work up between the independent machines, making them all function as one big computer. Most of the software for Linux/*nix computers is free, search for "Open Source clustering software", you are sure to find a lot on that topic.

Koko,

The USPS has a small distributed processing setup in every mail processing plant for image processing made up of 20-200 pizza boxes, each box being a quad core linux box with a master parceling out work based on response time and load monitoring. They communicate through a gigabit copper network.

Required result response time back to the originating sorting platform is 320 ms, and the system meets that mark 95-96% of the time. Our site is tiny, it's only set up to process about 150,000 images per hour. It's amazing what distributed processing can accomplish.

I'm well aware of the work that can be accomplished by distributed computer systems and clusters. I worked for GE Engineering Services for 25 years and we had a computer system called the DMC ( distributed Micro Computer) We had Hundreds of these all sharing work over a reflective memory network with Vax computer clusters suppling man machine interface and recipes. By your definition these would be supercomputers. These would be capacity computers or quasi-supercomputers.

My problems is the use of the term Supercomputer to describe a cluster of independent computers preforming complex task. These are Quasi-Supercomputers and do not come close to the true computing power of a true supercomputer that is a Capability computer manufactured by the likes of Cray, IBM, or HP. I guess the word Supercomputer is a little over worked in todays world because even Google uses a quasi-supercomputer for their search engine. I may be a little sensitive about this subject because I have worked around a
true Supercomputer and the one described in this writeup does not meet my criteria. Compared to some of the computers I have worked with, the PC I am posting with is a Supercomputer.

Here's a description of the old & the new Supercomputors being used
today.

http://www.purdue.edu/uns/x/2008a/08...ySICortex.html

While there is nothing new about the push for Green everything, Sometimes the added cost is hard to justify. The link below speaks for it's self.

The type of Computing offered by sicortex was more the quasi_supercomputer used for the information systems type of servers.

You might find this link of interest.
http://news.cnet.com/8301-11128_3-10250960-54.html