15 Dec, 2014 Industry Coverage
By Jeffrey Burt, December 15. 2015
The foundation’s president says the wins in 2014 will lead to more milestones next year, including new systems and the group’s first summit.
When IBM and several partners launched the OpenPower Foundation in 2013—and with it, the promise of Power-based scale-out systems built by companies other than Big Blue—the question arose as to whether the tech giant was too late to the market.
Cloud- and Web-based companies like Google, Facebook, Amazon and Microsoft—organizations that run huge scale-out data centers populated with massive numbers of servers to deal with such trends as mobile computing, big data, social networking and analytics—are having an increasingly larger influence on server and processor design and sales, due to the sheer number of systems they buy.
For these businesses, density, power efficiency and affordability are as important as performance, and often more so. It’s an area on which both Intel and ARM are focusing a lot of their efforts, and systems OEMs also are targeting. For example, Hewlett-Packard not only is building out its Moonshot portfolio of dense, low-power server modules, but also is teaming with contract manufacturer Foxconn to build Web-scale servers. The rise of hyperscale environments is also fueling the growth in the server market share by white-box makers like Tyan, Quanta and Wistron; IDC analysts earlier this month said that in the third quarter, original-design manufacturers (ODMs) had about 9 percent of server sales worldwide, a 44 percent jump over the same period in 2013.
It’s a crowded and increasingly competitive space, but IBM and its OpenPower partners are working to ensure that they’re heard above the din. The foundation has made significant strides since it was launched in August 2013—there are reference designs in the works, the embrace of accelerators and advanced networking in the Power architecture, Google testing its own Power-based motherboard—and 2015 promises even more advancements, according to Brad McCredie, an IBM Fellow, vice president of IBM Power systems and president of the OpenPower Foundation.
The group will hold its first OpenPower Summit March 17-19 in San Jose, Calif. In addition, more Power-based servers from vendors other than IBM will be hitting the market next year, McCredie told eWEEK. Since its founding, membership in the group has grown. On its Website, the group counts more than four dozen companies as members.
Such momentum should indicate to the industry that the OpenPower Foundation is not late to the game, he said.
“The growth that we’ve seen and the number of partners that have product plans with the architecture show that it is on time,” McCredie said. “It’s always better to do it early, but I don’t think it’s late.”
When IBM officials in April rolled out new Power8 systems and expanded on their goals for OpenPower, Patrick Moorhead, principal analyst with Moor Insights and Strategy, wrote a blog for Forbes praising the thrust of the group’s efforts but questioning whether IBM’s timing was off. However, Moorhead likes what he’s seen to this point.
“In the last year, OpenPower has made a lot of headway in developing its ecosystem compared to the ARM ecosystem and Intel ecosystem,” he told eWEEK. “They’ve grown quickly from zero to 60.”
Of key importance is the foundation’s adoption of accelerators, such as GPUs from Nvidia and field-programmable gate arrays from Altera and Xilinx, which plays to the demand within scale-out environments for heterogeneous infrastructures that are workload-optimized. The foundation also is putting efforts behind networking (with members like Mellanox Technologies) and memory (with Micron, Hynix and Samsung).
“We’ve opened up the platform to advanced networking and accelerators, a lot of what the warehouse guys [like Google and Facebook] want to do,” McCredie said.
The ecosystem is increasingly active, he said. There are between eight and 12 partners that are developing their own Power-based systems, and some already area showing off what they’ve created. Tyan in October unveiled the GN70-BP010 OpenPower single-socket customer reference system, and Google is running tests on its own two-socket board. Wistron officials also are talking about their upcoming two-socket Power system.
In addition, the program got a boost last month when the Department of Energy awarded IBM, Nvidia and Mellanox a $325 million contract to build two supercomputers for two national laboratories that will leverage technologies from the OpenPower Foundation.
It’s all good news for a group that understands the competition its faces.
“We know we’re going into a place where x86 is a dominant solution,” McCredie said.
Moorhead agreed. Intel has been aggressive in building out its capabilities for scale-out environments, not only with its low-power Atom platform but also through such technologies as its upcoming 14-nanometer Xeon D system-on-a-chip (SoC) and its Xeon Phi coprocessors. In addition, the company is integrating FGPAs into the same package as Xeon CPUs and growing its custom-chip capabilities.
It’s pushing back at the challenge by ARM and its partners—including Advanced Micro Devices, Applied Micro, Cavium and, most recently, Qualcomm—on the low end and, now, IBM and OpenPower on the higher end. It’s the system makers and their customers that stand to benefit. Not being Intel is an advantage for IBM, he said.
“People are looking for an alternative to Intel,” Moorhead said. “The industry likes to see competition out there.”
IBM’s McCredie said OpenPower is a growing option for OEMs and ODMs as the architecture challenges x86 in such important areas as price performance and energy efficiency and the ecosystem around it grows.
“There are a lot of things influencing the industry that will get them to [adopt] OpenPower,” he said. “People are looking for alternatives. People are looking for heterogeneity in the data center.”
For the full article go to eWeek