Science News

 

 Blue -Gene 'fastest super computer'

Article from The Computer Paper January 2000 Volume 13 No.1

NEW YORK-

IBM announced in early December that it has been working on the development of a new super computer that will represent a quantum leap forward in the way complex information is handled and processed.

The new super computer, named Blue Gene, will initially be used to understand and replicate how amino acids found in the human body intricately fold themselves into proteins. This, IBM believes, will lay the foundation for tremendous progress in the information intensive field of life sciences, and allow for the development of drugs that are not only targeted to combat specific diseases, but are further refined to provide maximum benefit for each individual patient.

Blue Gene will be able to

perform one million billion

operations every second,

and would be 500 times

faster than the world's

present fastest computer.

That would also make it

one million times faster

than the fastest desktop

currently in use
-Paul Horn, IBM

In opening the press conference called to announce Blue Gene, Paul Horn, IBM's senior vice-president for research, called the type of computing now being done by super computers situated in financial institutions, airline companies, and government offices, "deep computing." But, the capabilities of these machines pale in comparison to Blue Gene, which he described as a "Scream."

Horn said that Blue Gene, when fully operational in four to five, years, will be capable of speeds up to one "petaflop." In more simple terms, Horn said that blue Gene will be able to perform one million billion operations every second, and would be 500 times faster than the world's present fastest computer. That would also make it a million times faster than the fastest desktop currently in use, he said.

Ambush Goyal, IBM's vice-president for computer science, explained that to create a super computer with the capacity of Blue Gene, a whole new form of computer design, which he called "SMASH"' had to be created. SMASH, he explained, stood for simple, many, and self- healing.

The "simple" part of the design, Goyal explained, was in newly designed computer architecture that entailed simplifying the way computer chips are set up in today's computers, such as having memory and commands reside together on a chip.


"Many" refers to the fact that Blue Gene will have one million on-board processors, each capable of performing eight threads of execution at the same time. As an example of just how much

communication capacity was needed for these processors to "talk" with each other, Goyal said that, assuming sufficient bandwidth existed, the communications power resident in Blue Gene would be able to download all the pages on the World Wide Web in less than one second.


The "self-healing" aspects of Blue Gene refer to the need for the computer to detect, isolate, then solve any malfunction that may exist among the eight million total possible threads of execution. Goyal explained that this ability. was particularly important since, when fully operational and working on the amino-acid folding problem, it will be necessary for Blue Gene.to operate 24 hours a day, seven days a week for a full year in order to accomplish its objectives; it wouldn't be feasible for the machine to be shut down for troubleshooting.


To give some shape to the complexity of the project to understand how amino-acids fold into protein, Horn said that there were more possible ways of folding an amino-acid than there were atoms in the universe.
Horn did indicate that IBM would be partnering with medical institutions and pharmaceutical companies in the development of Blue Gene and the amino-acid project. Any of the expected financial contribution to building and maintaining Blue Gene, which he said IBM expected to receive from the drug companies, would be in addition to the US$100 million IBM expects to invest in developing Blue Gene.


But, Horn emphasized, Blue Gene would not be limited in function to only solving the puzzle of how amino-acids fold. It could also be used to simulate how the Big Bang occurred, or for the designing of "unbelievable graphic images."


According to Goyal, Blue Gene will be housed at IBM's Watson Research Center in Yorktown, N.Y., and will begin work relatively soon on analyzing small proteins, those which contain 30 to ·40 amino acids, as distinguished to the more common protein, which has about 300 amino acids.


Barely able to contain his excitement, Horn said, "This has got to be the best of times to be a researcher in information technology."

 Go Back To Science