If high-performance computing is becoming a mainstream part of corporate IT, then Sharan Kalwani, who manages HPC at General Motors, is its navigator.
Kalwani conducted a four-hour tutorial at the annual International Conference on High Performance Computing, Networking, Storage and Analysis, or SC06, with an eye toward giving HPC managers an overview of IT management basics. Among the topics covered were ROI, service-level agreements and portfolio management, with some homespun advice about what to expect from IT managers.
"They want the lowest-cost solution, and that's a battle that you find starting from Day One," Kalwani said to a class of about 60 people at the conference.
IT managers may also have trouble understanding some of the goals of people who use powerful and complex systems for research, said Kalwani. "IT, surprisingly -- despite the 'T' in IT -- are not technical. They're almost bureaucratic, so you've got to watch out for that," he warned.
After his tutorial, Kalwani also said that HPC users have to adapt. "HPC, now that's it has become mainstream, [should] also start acting like mainstream," said. He went on to say that the benefit of adopting IT processes will be "improved quality, definitely reduced costs and actually more wide acceptance."
Many of the people who attend SC06 live in a different world from mainstream IT. They are largely researchers from academia, government agencies and corporations. Many hold advanced degrees in mathematics, physics, biology and other sciences and use HPC technologies for everything from basic scientific research to business intelligence, product design and visualization.
This HPC culture, however, is increasingly coming under CIOs' management responsibilities as company officials look to control costs as well as integrate HPC processes into the business.
The problem with applying IT disciplines and measurements in an HPC research environment, said Goran Pocina, is figuring out how to adapt it "to a research community of users where quality isn't measured by how stable the environment is but on how quickly it can adapt and change."
Pocina, a technical adviser at a large pharmaceutical firm that he requested not be identified, said that his firm has individual research supercomputers installed globally that are managed locally by their own groups. These groups don't share applications, and they're not sharing processes. "The cost of maintaining this is tremendous," he said.
But if IT management could play a role, such as bringing in IT disciplines' best-of-breed applications and tools -- and getting rid of redundancies -- Pocina believes his company would benefit tremendously through improved productivity by scientists and reduced costs.
Irving Wladawsky-Berger, IBM's vice president of technical strategy and innovation, believes large businesses have crossed a dividing line separating HPC systems from IT. The use of these systems will only grow with their capabilities, he said. For their part, CIOs will have to learn more about these systems as their capabilities are used more often in commercial applications, he added.
"Traditional CIOs need more of the kinds of skills that before were only found in the HPC world," said Wladawsky-Berger.
Micah Nerren, an HPC consultant at Mach1 Computing, said he often works as a go-between for IT and HPC compute cluster groups that don't have the management skills or necessarily know how best to integrate their systems with the business. "You have to educate them a bit about how to coexist peacefully ... and educate IT [about] why this is a unique user," he said.