It's no secret that analytics is eating the enterprise world, but if there's anything in perpetually short supply, it's speed. Enter Cray, which on Tuesday unveiled a new supercomputing platform designed with that in mind.
Dubbed Urika-GX, the new system is the first agile analytics platform to fuse supercomputing with an open, enterprise framework, Cray said.
Due to be available in the third quarter, Urika-GX promises data scientists new levels of performance and the ability to find insight in massive data sets quickly. The system is tuned for highly iterative and interactive analytics, and integrated graph analytics offers rapid pattern matching.
"In the past, you'd run some types of analytics every 24 hours or even every week," said Ryan Waite, Cray's senior vice president of products. "Today, you might want to run them every six hours or every hour to be more in tune with what customers are doing."
Urika-GX enables concurrent workloads using Hadoop, Spark and graph technologies. Resources can be repurposed dynamically, and an open framework permits customization.
Featuring preintegrated, standards-based software, the system can be up and running in just a few days, Cray said.
"With some big data systems, you get a big pile of hardware and a recipe book," Waite explained. "You put it all together yourself, and the version interoperability issues can be crippling. What should take days could take weeks or months."
Overall, the new system is as much as three times faster than the previous Urika generation on complex queries.
Urika-GX is a standard 19-inch rack featuring industry-standard Intel Xeon processors, up to 22 TB of DRAM and as many as 1,728 cores per system. There's 35 TB of SSD storage and 192 TB of hard-drive storage per rack. It also taps the Cray Aries high-speed interconnect.
The markets for big data and high-performance computing are beginning to blur, said Nik Rouda, senior analyst with Enterprise Strategy Group.
"One trend that Cray can capitalize on is the movement toward appliances, or engineered systems for big data," Rouda said.
That's not currently the default model, but it's growing in popularity. Looking ahead, ESG predicts that an increasing proportion of the market will opt for this option as their primary deployment model because of benefits including performance, time-to-value and supportability, he said.
Cray's offering is also distinguished by its emphasis on high-speed networking, Rouda added.
Other vendors taking a similar approach include Oracle and Teradata. "The biggest challenge may be just building awareness for Cray’s offering in the broader market," Rouda said.