Leveraging the reputation of its high-powered Blue Gene supercomputer line, IBM Corp. Friday unveiled its Blue Gene on Demand offering to customers that want periodic ultra-high performance without having to own and maintain the hardware.

In an announcement, IBM said it will offer access to Blue Gene supercomputing power through its Deep Computing Capacity on Demand Center in Rochester, Minn. Enterprise customers can get extra computing power remotely when they need it, paying by the hour and number of processors used.

"Blue Gene is the kind of technology -- if it"s appropriate for your application -- that can really speed up a process," said David Gelardi, vice president of IBM"s Deep Computing Capacity on Demand program. The idea is that supercomputers can offer valuable computing power to businesses such as pharmaceutical firms, design businesses, financial institutions, weather modeling companies and others without their having to invest in new hardware.

The Blue Gene on Demand program will transmit data through a highly secure and dedicated virtual private network, according to IBM, and the data will be visible only to the company that owns it. Users pay only for the amount of capacity reserved and used on Blue Gene. Pricing will be approximately 50 cents per CPU per hour, Gelardi said.

"If you want to use Blue Gene, you would probably want to use a lot of it for only a short time," he said.

But Blue Gene on Demand isn"t for everyone, he said. The highly parallel, interconnected supercomputer requires software applications that are specifically tailored for its architecture. "Finding the match between the applications and Blue Gene is really the trick," Gelardi said.

Typical users will probably include bioinformatics, quantum chemistry, astronomy and space research groups. "We know that the characteristics of those applications will fit well within the confines of Blue Gene," Gelardi said.

The company expects other business needs will follow, including business and financial analytics.

Lance Westerhoff, chief software engineer at QuantumBio Inc., a pharmaceutical software company in State College, Pa., said his company has been working with IBM to allow customers to use its drug-dependency analysis software on Blue Gene. By automating much of a drug company"s drug testing using ultra-fast computers like Blue Gene, companies can reduce costs and time to market for new drugs, he said.

He said drug companies could also reduce the amount of costly laboratory testing by zeroing in on the best possible drug candidates and only doing real-world lab tests on those. Typical drug discovery processes take 10 to 15 years today, which could be shortened considerably using supercomputers, he said.

But convincing drug companies of the benefits of using Blue Gene means first convincing them that the system -- and their data -- is absolutely safe, he said. "The major concern of the pharmaceutical industry is security," Westerhoff said. "They don"t like sending drug information off that someone else could use. If they would send their data to other people and if the security can"t be guaranteed, they"re not interested."

Westerhoff said the benefits will eventually win out over the security fears. "The customers are really ripe for it right now," he said of Blue Gene"s computing capabilities. "Everybody knows they need to get things done much faster."

Vernon Turner, an analyst at IDC in Framingham, Mass., called IBM"s marketing of Blue Gene power to commercial businesses intriguing.

"Using the power of the brand, of the Blue Gene name, and bringing it into the enterprise would be something of interest to a number of customers," including financial services companies, Turner said. "The fact that they"re making the Blue Gene capabilities available to the commercial market rather than just the scientific market is the interesting piece."

The Blue Gene on Demand program uses IBM"s eServer Blue Gene machines, which became available commercially in November. Each eServer Blue Gene machine is available in up to 64 racks, with 1,024 dual CPU nodes per rack. The peak theoretical performance is 5.7 TFLOPS per rack, according to IBM.

The Rochester Blue Gene on Demand Center includes more than 2,000 IBM PowerPC CPUs running on Linux. In addition, IBM has capacity-on-demand centers in Poughkeepsie, N.Y., and Houston, as well as in Montpellier, France. In those centers, customers have access to more than 5,200 CPUs to run Linux, Microsoft Windows and IBM AIX operating environments.

Blue Gene supercomputers have been used for years in national laboratories and other institutions for defense, weather and other research. IBM"s Blue Gene/L supercomputer at the Lawrence Livermore Laboratory in Livermore, Calif., has been running a 32,000-CPU system since December, and IBM is in the process of doubling its size to 64,000 CPUs.