HP to spin its R&D future with The Machine
June 11, 2014
Calling it a mission HP must accomplish because it has no other choice, HP Labs director Martin Fink is announcing a new computer architecture Hewlett-Packard will release within two years or bust. Fink, who was chief of the company's Business Critical Systems unit before being handed the Labs job in 2012, is devoting 75 percent of HP's Labs resources to creating a computer architecture, the first since the company built the Itanium chip family design with Intel during the 1990s.
A BusinessWeek article by Ashlee Vance says the product will utilitize HP breakthroughs in memory (memsistors) and a process to pass data using light, rather than the nanoscopic copper traces employed in today's chips. Fink came to CEO Meg Whitman with the ideal, then convinced her to increase his budget.
Fink and his colleagues decided to pitch Whitman on the idea of assembling all this technology to form the Machine. During a two-hour presentation held a year and a half ago, they laid out how the computer might work, its benefits, and the expectation that about 75 percent of HP Labs personnel would be dedicated to this one project. “At the end, Meg turned to [Chief Financial Officer] Cathie Lesjak and said, ‘Find them more money,’” says John Sontag, the vice president of systems research at HP, who attended the meeting and is in charge of bringing the Machine to life. "People in Labs see this as a once-in-a-lifetime opportunity."
Fast, cheap, persistent memory is at the heart of what HP hopes to change about computing. In the effort to build The Machine, however, the vendor harks back to days when computer makers created their own technology in R&D organizations as a competitive advantage. Commodity engineering can't cross the Big Data gap created by the Internet of Things, HP said at Discover today. The first RISC designs for HP computers, launched in a project called Spectrum, were the last such creation that touched HP's MPE servers.
Itanium never made it to MPE capability. Or perhaps put another way, the environment that drives the 3000-using business never got the renovation which it deserved to use the Intel-HP created architecture. Since The Machine is coming from HP's Labs, it's likely to have little to do with MPE, an OS the vendor walked away from in 2010. The Machine might have an impact on migration targets, but HP wants to change the way computing is considered, away from OS-based strategies. But even that dream is tempered by the reality that The Machine is going to need operating systems -- ones that HP is building.
OS compatibility was one reason that Itanium project didn't pan out the way HP and Intel hoped, of course. By the end of the last decade, Itanium had carved out a place as a specialized product for HP's own environments, as well as an architecture subsidized by Fink's plans to pay Intel to keep developing it. The Machine seems to be reaching for the same kind of "change the world's computing" impact that HP and Intel dreamed about with what it once called the Tahoe project. In a 74-year timeline of HP innovation alongside the BusinessWeek article, those dreams have been revised toward reality.
The Machine, should it ever come to the HP product line, would arrive in three to six years, according to the BusinessWeek interview, and Fink isn't being specific about delivery. But with the same chutzpah he displayed in running Business Critical Systems into critical headwinds of sales and customer retention, he believes HP is the best place for tech talent to try to remake computing architecture.
According to the article, three operating systems are in design to use the architecture, one open-source and HP proprietary, another a variant of Linux, and a third based on Android for mobile dreams. That's the same number of OS versions HP supported for its first line of computers -- RTE for real time, MPE for business, and HP-UX for engineering, and later business. OS design, once an HP staple, need to reach much higher to meet the potential for new memory -- in the same way that MPE XL made use of innovative memory in PA-RISC.
Fink says these projects have burnished HP’s reputation among engineers and helped its recruiting. “If you want to really rethink computing architecture, we’re the only game in town now,” he said. Greg Papadopoulos, a partner at the venture capital firm New Enterprise Associates, warns that the OS development alone will be a massive effort. “Operating systems have not been taught what to do with all of this memory, and HP will have to get very creative,” he says. “Things like the chips from Intel just never anticipated this.”