Turbocharged computers could soon be a boon to system builders. In fact, the market for high-performance computers used by financial firms, engineering companies, and universities will grow 7 percent annually over the next few years, according to market analyst firm Intersect360 Research. That translates to $36 billion in software, hardware, and services by 2015.
So the market could represent friendly terrain for custom builders and resellers. And, multiple players are looking to leverage the channel. But high-performance computing requires a deep understanding of what customers are trying to accomplish. It also requires a look ahead to problems as yet unsolved. “Once you’ve designed a bridge, you don’t need to design that bridge again,” says Christopher Willard, chief research officer at Intersect360 Research. “You go and find the next bridge you need to design.”
As result, system turnover averages three years and applications are updated frequently. Further, there’s not much of a drag from legacy systems, and customers are willing to work with builders to get to the next performance level on budget. These parameters also mean that customers can be fickle, however. What’s more, acceptance testing can take months and payment can be slow.
Bruce Toal, CEO of Richardson, Texas-based startup Convey Computer Corp., says his company’s reconfigurable coprocessor can accelerate the solving of problems that have a high degree of parallelism up to fortyfold. Toal is looking for specialized, knowledgeable, and customer-focused resellers who can help in such emerging areas as using social media to ferret out brand passion analytics. “There’s tremendous, timely information available if you can sift through it quickly,” says Toal.Panasas is looking for help from the channel after largely using a direct sales model. The Sunnyvale, Calif.-based company makes storage systems for high-performance computing. Inadequate storage response can lead to systems being starved for data. Slow data movement can be particularly vexing in some situations, says Chief Marketing Officer Barbara Murphy. “Financial institutions are trying to execute on some problem that they want to solve,” she says, “but the timeline to solve the problem has to fit within a day trade.” Murphy sees tremendous opportunities for resellers in oil and gas exploration, the company’s core market.
High-performance computer maker Appro International Inc., of Milpitas, Calif., has partnered with Panasas for three years, says Anthony Kenisky, Appro’s vice president of North America sales. He notes that growth in the high-performance market has been driven in part by the acceptance of commodity components, which are inexpensive and now offer the required performance per watt. However, in systems with thousands of nodes that sometimes run for weeks, there will be disruptions or failures. “We have to account for that and address those via the software stack,” Kenisky says, adding that this capability sets Appro apart.
Another high-performance computing player turning to the channel is PSSC Labs, based in Lake Forest, Calif. Alex Lesser, vice president of sales and marketing, says the company is moving toward the concept of a black-box supercomputing appliance for such specialized applications as drug discovery, molecular modeling, astrophysics rendering, and others. “We want to bring our solutions to a more mass [market],” Lesser says. The company is in discussions with a distributor to do just that. A deal should be announced in 2012, he notes.
Two years ago, Fremont, Calif.-based Penguin Computing launched an on-demand high-performance computing service that targets the engineering market, according to CEO Charlie Wuischpard. There is the possibility, says Wuischpard, that the service could be branded and sold by VARs.
Wuischpard sees many high-performance computing opportunities, driven in part by the growing stacks of data with which companies, universities, and others want to work. There’s plenty of room for small players, but only if they’re knowledgeable about customers’ problems and can offer possible solutions. As Wuischpard says, “The challenge is still going to be, do you understand the applications and how to put them together, especially the larger cluster solutions?”