By Charles King, Pund-IT, Inc. November 15, 2015
Anyone who follows the supercomputing/high performance computing (HPC) space knows that for all of their jaw-dropping capabilities these systems follow the same populist path familiar to every other technology. That is, thanks to ongoing technological innovation everything that was once exceptional and usually unaffordable becomes commonplace and available to increasing numbers of users and organizations.
But a big question to consider is how long this process usually takes and, even larger, whether anything can be done to effectively speed the process. Those points are of particular interest to Intel and a central goal in the company’s Scalable System Framework (SSF) effort, including the announcements the company made at Supercomputing 2015 (SC15) this week.
Intel calls SSF an advanced architectural approach that enables more scalable, flexible and balanced HPC solutions. In essence, the company is working to create high performance systems and system designs whose compute, memory and I/O components can be efficiently managed and consistently balanced to rapidly bring HPC-class benefits to commercial workloads.
Intel believes that this will allow HPC systems to enrich a wide variety of common business use cases, like big data analytics and cloud, and help to usher in what the company calls an era of “HPC everywhere.” At SC15, Colfax, Cray, Dell, Fujitsu Systems Europe, HPE, Inspur, Lenovo, Penguin Computing, SGI, Sugon and Supermicro all announced plans to launch systems based on the Intel SSF early next year.
Intel’s other SC15 announcements ran the gamut from planned architectures, designs and validation tools to a new power-efficient end-to-end fabric solution (Intel OPA) that can support larger HPC clusters than commonly used Infiniband EDR fabrics. Intel said that the OPA fabric is already being used at a number of high profile HPC sites.
The company also noted that preproduction versions of its Xeon Phi (Knight’s Landing) processors are up and running in several supercomputer class systems. Finally, Intel is one of 30+ founding members of the OpenHPC Collaborative Project which is developing a comprehensive and cohesive open source system software stack to drive broader adoption of HPC.
So how does the news Intel shared at SC15 stack-up against the company’s larger goals for HPC and supercomputing? Pretty impressively.
Intel’s Xeon Phi solutions for highly parallel processing are progressing well commercially, and the new Intel OPA technology could significantly alter the HPC fabric status quo. Along with its standalone efforts, the company is also demonstrating (via the OpenHPC Collaborative Project) how well it can play with others, a critical point in a sector where multiple vendors participating in system design/deployment is commonplace. Most intriguingly, Intel’s SSF architectural efforts could be even more disruptive in terms of accelerating the commercial adoption of HPC.
That would benefit numerous industries and organizations but it could also shift the economics of HPC in ways that are favorable to vendors. Traditionally, supercomputer access has been limited mostly to deep-pocketed government labs and associated universities. But development deals typically require supercomputer vendors to initially subsidize system development and to wait months or even years to be reimbursed. If SSF delivers as Intel is planning, it could provide a balm for those painful upfront investments.
Finally, it’s important to consider the fact that what Intel is doing in HPC and supercomputing is simply the latest of numerous company efforts designed to transform computing and the IT industry. Intel’s focus on developing leading edge industry standard technologies and components, and then using its sophisticated manufacturing processes to deliver them to market has fundamentally changed the way businesses and consumers utilize and benefit from IT.
In short, Intel has done it before and there is every likelihood that offerings, including SSF, Intel OPA and Xeon Phi, along with efforts like the OpenHPC Collaborative Project will help them do it again. If that is correct, we may well be on the cusp of the era of HPC everywhere.
© 2015 Pund-IT, Inc. All rights reserved.