Big Data // Hardware/Architectures
News
5/7/2013
08:12 AM
Connect Directly
Google+
LinkedIn
Twitter
RSS
E-Mail
50%
50%
Repost This

Cray Builds A Budget-Minded Supercomputer

Cray introduces entry-level supercomputer starting at $500,000 for high-end engineering, computational modeling and simulation work.

Interop 2013 Las Vegas Collage: 10 Cool Products at Interop
10 Cool Products at Interop
(click image for larger view and for slideshow)

Well-known supercomputer vendor Cray introduced what it describes an "aggressively priced" product line on Tuesday aimed at an emerging class of customers delving into high-performance computing (HPC) perhaps for the first time.

Cray's top-end models used by large corporations, big government research centers and university labs require raised-floor data centers, liquid cooling, optical networking cables and run upwards of $10 million to $20 million for a complete system. The new Cray XC30-AC supercomputer uses many of the same components and software elements as those monsters, but it is air cooled, requires no optical cables and costs from $500,000 to $3 million. The target market is Fortune 100 to Fortune1000 manufacturers, pharmaceutical companies, and oil and gas firms, and smaller universities, government agencies and research labs that want to exploit supercomputing techniques.

"We're seeing a new wave of demand because companies are gathering more data and asking more complex questions than ever before," said Barry Bolding, Cray's VP of storage & data management and corporate marketing, in an interview with InformationWeek.

Aerospace Giants like Boeing, GE, and Pratt and Whitney are no strangers to HPC, but now smaller aircraft and jet engine manufacturers, for example, are among the types of firms looking at supercomputing to model airflows and eke out that much more efficiency in their products, said Bolding. Auto, truck and air foil manufacturers are doing the same to improve the fuel efficiency of vehicles while smaller petroleum exploration and pharmaceutical firms are doing high-end imaging and modeling work.

[ Want to hear about IBM's $3.3 million HPC option? Read Rutgers Gets Big Data Weapon In IBM Supercomputer. ]

These types of problems and analyses aren't addressed by relational database systems or alternative processing platforms such as Hadoop. Smaller customers are experimenting through time sharing and hosting arrangements, but firms intent on cutting their development cycles are bringing HPC systems in house.

Cray is counting on XC30-AC filling a niche between its own top-end products and those of long-time competitor IBM at the high end, and what Bolding calls "commodity cluster" competition from the likes of Dell, HP and SGI at the low end. Like the latter crowd, Cray uses chips from AMD, Inteland Nvidia (having ceased manufacturing of its own chips seven years ago), but it maintains performance advantages through its internal networking and connection technologies, according to Bolding.

"When you're connecting thousands of cores as we are, it gives you a very large performance advantage," Bolding said. "Our interconnect is 10% to 20% faster than Infiniband on point-to-point connections and up to 10 times faster on total-system latencies."

Beta customers for the XC30-AC include a consumer electronics company and a global financial services firm. Neither company was identified, but both were described as the types of non-traditional Cray customers that are venturing into high-performance computing.

The XC30-AC is available immediately and comes with a Linux-based operating system and software environment including a compiler and software tools for parallel code.

Advances in chip design keep improving the performance of commodity clusters, but Cray has managed to keep its specs ahead of those competitors because "scaling complex problems is a harder problem than commodity system vendors can afford to keep up with," Bolding said.

"Cray has a history of leveraging the supercomputing technologies featured in its high-end systems, and economically repackaging those same technologies to offer solutions to fit the needs of HPC users with smaller budgets," said IDC HPC analyst Earl Joseph in a statement.

Comment  | 
Print  | 
More Insights
Comments
Newest First  |  Oldest First  |  Threaded View
Andrew Hornback
50%
50%
Andrew Hornback,
User Rank: Apprentice
5/8/2013 | 4:15:19 AM
re: Cray Builds A Budget-Minded Supercomputer
One would have to think that by going after smaller clients and providing compatibility between the smaller and larger systems that Cray is looking to capitalize on organizations looking to increase their ROI with regards to development costs that will eventually move to the larger systems.

The idea of making HPC more accessible to smaller educational institutions is an excellent move and something that I wish had happened 20 years ago. I'd have loved to have worked on a Cray system when I was in school.

Andrew Hornback
InformationWeek Contributor
In A Fever For Big Data
In A Fever For Big Data
Healthcare orgs are relentlessly accumulating data, and a growing array of tools are becoming available to manage it.
Register for InformationWeek Newsletters
White Papers
Current Issue
InformationWeek Government, May 2014
NIST's cyber-security framework gives critical-infrastructure operators a new tool to assess readiness. But will operators put this voluntary framework to work?
Video
Slideshows
Twitter Feed
Audio Interviews
Archived Audio Interviews
GE is a leader in combining connected devices and advanced analytics in pursuit of practical goals like less downtime, lower operating costs, and higher throughput. At GIO Power & Water, CIO Jim Fowler is part of the team exploring how to apply these techniques to some of the world's essential infrastructure, from power plants to water treatment systems. Join us, and bring your questions, as we talk about what's ahead.