Cray doesn’t do Top500 press releases says Barry Bolding, Cray SVP and Chief Strategy Officer. Not a bad strategy when you don’t have to: five Crays are in the Top Ten of the latest Top500 list (Nov. 2015) including the two newest machines (Trinity (DOE) and Hazel-Hen (HLRS)). Media coverage and market notice are inevitable.
Indeed Cray’s success at the supercomputing pinnacle is a thing of envy. It has helped drive roughly 20 percent annual revenue growth and near tripling of the company size over the past few years according to Bolding. Cray not only missed the recent supercomputer market softening but also won its share of the next generation, big machine procurements in the race towards exascale.
Cray didn’t deliver much new product news at SC15 agrees Bolding. Next year, he emphasizes, will be quite different. Nevertheless Cray was busy, hammering its supercomputers-must-be-productive-not-just-fast message and touting its expanding support of tools to do just that. Of particular note are the rise of big data challenges and efforts to turn HPC into a competitive weapon more accessible to a larger enterprise market.
Cray did, for example, announce an HPC-optimized Docker solution for the XC line of supercomputers and will extend that to CCS400, XE and CXK platforms in 2016. The intent is to provide better application portability and bring ease-of-use to running highly-scalable applications on advanced supercomputers.
With Docker as part of Cray’s official software distribution, Cray’s customers will be able to implement software containers that package an entire software stack — application codes, shared libraries, base operating system files, user environment variables, dependencies and more — into easily-deployable images that contain everything an application needs to run with increased efficiency across multiple platforms.
“[Improving] HPC productivity is something we want to see more of at conferences like this,” said Bolding, citing increased SC attendance by commercial entities and their laser focus on driving bottom lines higher.
“Clearly this convergence of big data and supercomputing is going to have a big impact. What you see are technologies migrating between analytics and HPC. We’ve had a series of products we’ve announced of the last couple of years (e.g. URIKA-GD) where we are bringing HPC technologies to the analytics space, first for very large graph problems and we’ll be continuing to do that. We are going to be continuing to bring HPC technologies to analytics problems,” he said.
HPCwire managing editor John Russell sat down with Bolding at SC15 for a brief but wide-ranging conversation looking at important trends in 2016. Besides the competitiveness imperative, topics included the OpenHPC initiative, new HPC ROI metrics (thank you IDC), and NSCI.