Cray Forecasts More Enterprise Contracts
Global supercomputer maker Cray's recent deals with Danish and Swiss meteorological organizations further cemented its leadership in the weather business, but Cray predicts blue skies ahead for expanding sales across an diversified group of corporate clients.
The company this week inked a contract with longtime customer Swiss National Supercomputing Centre (CSCS), which installed a Cray CS-Storm cluster supercomputer to run the Swiss Federal Office of Meteorology and Climatology's (MeteoSwiss) operational numerical weather forecasts. In late August, the company the Danish Meteorological Society awarded Cray (NASDAQ-CRAY) a $6-million contract for a Cray XC supercomputer and Cray Sonexion 2000 storage system.
The Swiss contract uses a GPU-accelerated supercomputer to power production numerical weather models for its national weather service, the first time this has been done, according to Cray. But accomplishing this feat was critical to MeteoSwiss, which had worked on the challenge internally, said Barry Bolding, senior vice president and chief strategy officer at Cray, in an interview.
GPUs are applicable for a smaller set, but typically have to do more work to get an application to work and scale, he said.
"If you can do that, you potentially have something more compact and more energy-efficient," said Bolding. "Most of the performance in the system is being derived from these high-performance Tesla processors from Nvidia. They can do more work for less power; power is important. You can get more work done per kilowatt of electricity that you draw into your datacenter. Because the GPUs are very fast at doing computation, they can do a finer resolution model. The Swiss needed a finer resolution model because of the mountainous terrain they have. Within their datacenter, they can do a computation that's 40-times more powerful and scale the resolution that’s a factor of two."
This flexibility, scalability, and power are reasons Cray expects continued growth in the enterprise market as well as its traditional markets. With some systems, like the CF Storm, built to scale up to 30 racks, and others like the GPU designed for some smaller configurations, Cray increasingly sees commercial success at enterprise customers, Per Nyberg, senior director of worldwide business development at Cray, told EnterpriseTech.
Faced with a never-ending influx of data and the mandate to empower business users to translate this flood into usable insight, CIOs more frequently recognize traditional enterprise approaches are inadequate or cost-prohibitive, said Nyberg. And more enterprises are turning to Cray as a direct result, said Bolding.
"There's definitely a deluge of data that is causing increased interest in Cray as the company to help them deal with the data and do heavy duty, scalable analytics and heavy duty scalable insights into that data," he said. "There's a big data problem now. It's driving a bigger computer to do the simulation and the modeling."
Blending a supercomputer approach with traditional enterprise IT provides enterprises with the data movement conceptualization they need, coupled with scalable, powerful, and reliable technologies, added Nyberg.
"With this influx of data, it's breaking the infrastructure they have. [CIOs say], 'We fully recognize that what we have today kind of works or satisfies the minimum of what we need to do but we're already starting to see it break and we're not sure how to fix it," he said. "Just about every industry is starting to look at supercomputing and analytics."
The conversation differs depending on the corporate point of contact, said Bolding. CIOs and chief technology officers want to discuss the total cost of ownership and scalability, he said.
"They want to make sure you're also providing good value over time. When Intel comes out with a new processor … for a fraction of the cost, we can bring it up to the latest technology. We're allowing you to keep that infrastructure for a very long time because we'll pop out the old processor and pop in the new processor," said Bolding.
"We also appeal to the geeks at our customers – the application programmer – who's trying to get the most out of their technology. We're a company full of technological geeks so we also speak to the commercial end-users at the organization. At some companies you need to get the buy-in of the CIO. At other sites they defer to their technical specialists."
Although the Internet of Things is only expected to dramatically inflate the amount of data available to organizations for analysis, to date little of that data has been tapped for simulation. As a result, it has yet to affect Cray – or its bottom line, said Bolding. But that will change, no doubt.
"There's a huge piece of that hasn't yet impacted Cray and Cray hasn't yet had had an impact on. We're not integrating with Nest yet. Down the road, who knows?" he said. "As a city or large-scale government tries to manage its energy efficiently, those become the types of problems Cray's infrastructure can help with."