Hewlett-Packard is going for broke with a superbly ambitious project that aims to do nothing less than reinvent computer architecture with non-volatile memory technology. The initial target of its project seems to be servers, but the ultimate plan is for the tech to trickle down into all computer devices, which could one day mean Android smartphones packing a whopping 100TB of storage.
Critics might see the project as a desperate gamble for beleaguered HP, which, like other tech giants, is seeing revenues slip away from its traditional hardware businesses. But HP is deadly serious nonetheless, announcing at this week’s HP Discover event in Las Vegas that it’s throwing almost three-quarters of its research team at the effort, which might take anything from three years to a decade to see the light of day.
During an keynote speech covered exclusively by theCUBE, HP said that what it’s trying to do is build a new computer system from scratch, cobbling together a bunch of highly advanced yet unproven technologies to create “The Machine”. These technologies include something called ‘Memristor’, a long-delayed memory substrate that’s straight out of HP’s own labs, plus silicon photonics, customized processors and a brand new operating system built from scratch.
“HP is building a new way to compute from the ground up,” HP CEO Meg Whitman proclaimed in a keynote speech yesterday. “We’ve been talking about many of the component technologies for some time. Now we’re bringing them together in a single project to make a new computer architecture available by the end of the decade.”
HP’s new architecture can be whittled down to three simple concepts, namely: Electrons Compute; Photons communicate; and Ions store. The most important piece of the puzzle is HP’s memristor, a technology its researchers have been fumbling with for years. Memristor is a new kind of resistor that can store data even after the power source is cut off. It can potentially deliver massive amounts of super fast storage with minimal power consumption, making it highly desirable for HP’s ‘Machine’.
Memristor is crucial, because traditional computer architecture that uses DRAM and Flash is unable to keep up with the amount of data people are using, notes Businessweek. With memristor, IT would be able to do away with the old slow disk/fast memory system seen in today’s computers, allowing the processor to access data far faster than it’s able to do now.
HP’s biggest obstacle will be what it’s struggled to do all along. That is, making the stuff. The tech was actually conceived way back in the early 2000s, but HP’s been unable to hone the process of mass manufacturing it. It first promised to release the tech as a replacement for flash in 2013, before delaying it until 2014, and later, 2018, notes The Register. Now though, HP’s done a bit of a turnaround and claims it’s ready to introduce the tech by 2016 according to a slide shown at the keynote.
The big question is whether HP really can deliver on this promise, especially after so many delays already? It’s not just manufacturing issues HP has to contend with, but also the cost of doing so, notes Wikibon analyst Dave Vellante. “Can HP get the costs out of memristor and compete with dram and flash – which have consumer volumes already?” he asked.
Silicon photonics and open-source tech
The other components are also a big deal. Silicon photonics provides an alternative to fiber-optic technology, which the likes of HP, Intel and others have struggled to squeeze inside computers, let alone a smartphone. Silicon photonics allows extremely high-speed, low latency data transfer, and could potentially replace Ethernet cables and be used to link whole racks of servers together.
“This will enable us to deal with massive, massive datasets [and] not just be able to take those massive datasets, but ingest them, store them, and manipulate them, and do this at orders of magnitude less energy per bits, per compute,” said Martin Fink, HP Labs director and CTO.
Of course, all of this new memory and networking technology will need an operating system to run on, and sadly for Microsoft that won’t be Windows. The problem is that most programs and apps are designed to run on today’s architecture, and assume that the memory systems feeding the processor are slow. That’s why HP has assigned a separate team of researchers to work on a new, open-source Machine OS that’s designed to run on machines with a high-speed, constant memory store. Open-source is of course a crucial part of HP’s strategy, as we’ve seen with its earlier forays into OpenStack, and so it’s no surprise it’s looking to open-source its OS as well.
Besides this, it also has open-source Linux and Android operating systems in the works, presumably in anticipation of the day its new architecture can be squeezed into everyday devices like PCs and phones.
“We want to reignite in all of our universities around the world operating system research, which we think has been dormant and stagnant for decades,” added Fink.
Can HP cobble it together?
The Machine is an entirely conceptual one at the moment. For now, the components will all be developed and sold independently, and it’s by no means certain that HP can overcome the enormous technological challenges it’s taking on. Nevertheless, as Dave Vellante points out on theCUBE, it’s encouraging that the company is at last getting back to its roots. “It makes HP relevant for the future, and it gets them back to their roots of invent,” he stated.
*see more analysis from Dave Vellante here:
It’ll be several years before we see if HP can deliver on its promise and reinvent the computer, but if it can it will surely secure its future for many years to come. After all, HP is a company that was founded on a high-margin business, but those have floundered with the rise of Cloud and white-box hardware makers. HP’s going to need a new high-margin business if its to survive, and ‘The Machine’ might just be it.