HP & The Machine: Big Hopes, Big Hurdles

10 More Powerful Facts About Big Data

Even the name sounds awesome: The Machine.

It's not a movie coming soon to a theater near you. Rather, it's an ambitious framework for the future of computing recently announced by HP -- one that is not coming to an office or datacenter near you anytime soon, at least not in the exact form HP has laid out. The company said its timeline for The Machine runs through 2020, but even that might be aggressive. In the meantime, The Machine makes for compelling discussion. HP, for its part, is saying the right things, according to Forrester VP and principal analyst Richard Fichera.

"The guiding principles they're talking about are all wonderful," Fichera, who served as director of BladeSystem strategy at HP from 2006 to 2010, said in an interview. There's a catch, though: The Machine, no matter how good it sounds, faces a long road to becoming everyday reality. "This is all great stuff. However, not much of it is really under [HP's] control."

Fichera gave high marks for the ideas behind The Machine and its attempt to address rapidly growing data and the demands it will place on IT departments, datacenters, networks, and other infrastructure. Nor is the marketplace need for such innovation absent.

[It's all about speed. Read IT Innovation At Full Speed: InformationWeek Video.]

"The things they're talking about are indeed the right kind of things to be thinking about for building massively, orders-of-magnitude more scalable systems than we have today -- and there's a legitimate demand for it," Fichera said. He offered, as an example, conversations he's had with IT pros at financial services firms, who say they have a pipeline of business application requests that will require processing power of two-to-three decimal orders of magnitude more than they can afford today: "The demand is there, the vision is the right vision."

That said, a host of challenges create a pretty wide gap between vision and execution, according to Fichera. This becomes more apparent when looking at HP's four key components of The Machine: memristors, system on a chip (SOC), photonics, and operating system.

Memristors, for instance, aren't new in the research-and-development world. While the concept behind them is on target, Fichera noted that they're still a long way from commercialization and they face competition from other technologies.

Figure 1:

Meanwhile, he continued, "[SOC is] another great idea, but one that HP's probably not going to do by themselves. They have some influence because they're one of the world's largest purchasers of processors." HP's Moonshot microservers, for instance, are "really good technology," according to Fichera, but they rely on microprocessors and SOC technology from other firms. Expect Intel, AMD, and others to be factors on the SOC front. "HP is not building the SOC processor and SOCs anymore -- not for a long time," Fichera said. "HP will bring some design expertise to build servers using SOC parts, but so will everyone else in the business -- everyone's looking at this."

On the photonics front, Fichera said HP has done a lot of good research and owns an impressive patent portfolio. Yet the implementation as envisioned for The Machine isn't entirely in HP's hands. "The level they're talking about -- eliminating copper wires, especially inside the system enclosure -- is, again, going to be driven probably by Intel because

they're the volume semiconductor supplier. Photonics isn't going to realize its full potential until Intel starts producing parts that have photonic inter-connects, and that's out on the horizon." Fichera added that the "engineering jury" hasn't yet decided the right place for photonics to reside, either.

As for operating system: "HP doesn't write OSs. They write cloud environment software right now, and they're enhancing their own and they're enhancing OpenStack," Fichera said. "Just saying we need a new OS isn't going to make one magically appear."

To be clear, this isn't all to naysay the concepts and market demands driving The Machine. "Their vision is correct: these are all the kinds of technology that will be needed," Fichera noted. Rather, the ambitious nature of the program means even a behemoth like HP will need some help -- lots of it -- to turn even a revised version of that vision into reality. Fichera noted that 2020 might sound far away, but it's not. "That's not a lot of time to build an entirely brand-new architecture," he observed.

The more reasonable bet is that this kind of thinking will push HP and other firms to better meet the needs of enterprises wrangling with big data, cloud, and other needs on so-called legacy systems. In other words, we may not see The Machine arrive in 2020 -- or ever, for that matter -- as currently described, but we will see innovation that turns some of The Machine's theoretical upside into reality.

"Whether or not they get 'The Machine,' what this kind of thinking guarantees is that they will get progressively better systems to solve the problem they described," Fichera said. For example, the advent of commercialized memristors could "essentially result in having computers that have a single-level store below main memory -- that there's no longer any differentiation between local disk, network disk, flash, and so on. It will be all one persistent, fast, and very big second-level store."

The bottom line: "There's a ton of people that need to buy into this [for The Machine to become reality]," Fichera said. HP may be overstating its own role and capability in delivering something like The Machine, but that doesn't mean there's no upside.

"HP's in a very good position to deliver systems that utilize this kind of architectural thought," Fichera said. "[But] there's no way anyone can deliver everything they've got here in a working system in six years. That's asking a lot. But the net fallout will be we will get much better systems as more and more people look at these kinds of contributing technologies. If we change HP's recipe to 'system on a chip, a better persistent memory layer, better inter-connection and data transmission, and improvements in operating system and software environment,' those four things are kind of defining what we need to do to solve this problem."

InformationWeek's new Must Reads is a compendium of our best recent coverage of the Internet of Things. Find out the way in which an aging workforce will drive progress on the Internet of Things, why the IoT isn't as scary as some folks seem to think, how connected machines will change the supply chain, and more. (Free registration required.)