IBM and genetics research firm NuTec Sciences Inc. today announced that they're building a Unix-based clustered system that the two companies claim will be the largest supercomputer installed to date by a corporate user.
The cluster will tie together 1,250 of IBM's eServer p640 systems with a combined total of 2.5TB of memory and 50TB of online disk storage, according to officials at the computer maker and Stafford, Texas-based NuTec. The research firm's NuTec Life Sciences division in Atlanta will rent time on the system to pharmaceutical and biotechnology researchers looking to investigate how genes interact in the human body to cause life-threatening diseases.
A computer algorithm created by the National Human Genome Research Institute -- a U.S. government agency that's part of the National Institutes of Health (NIH) -- lets researchers look at the way multiple genes combine to create diseases, said Pete Morrissey, president of NuTec Life Sciences.
"Very few diseases, maybe 2%, are caused by a single gene," Morrissey said. But a limiting factor affecting the use of the NIH algorithm in research applications has been the massive need for computing speed and power in order to process all of the relevant genetic data, he added.
NuTec's supercomputer is supposed to be capable of processing 7.5 trillion calculations per second, said James Coffin, head of a division at IBM that targets life sciences applications. That would rank the system among the 10 fastest computers in the world and make it the most powerful one outside of government laboratories, Coffin said.
Hardware vendors such as IBM and Compaq Computer Corp. are pairing up with bioscience companies to tackle big genetics-related processing jobs. For example, researchers at The Celera Genomics Group in Rockville, Md., used a supercomputer built by Compaq to power through the basic sequencing of the human genome structure earlier this year.
But there's more to finding cures to diseases than the application of computer horsepower to genetic data, Morrissey said. It's also "critically important that we find a way to make visualization tools to make the data make sense" to researchers, he said. And researchers have to figure out a way to use the machines as efficiently as possible, Morrissey added.