HP is trying to solve the big data problem, just like all of their competitors. The company has slowly been preparing the building blocks for their strategy to attack big data with things like HP’s Moonshot. Many believed that HP’s Moonshot was the company’s sole hardware answer to the big data problem, but in reality it was merely a piece of what HP is calling The Machine. Essentially, The Machine combines three technologies that HP is developing or has already developed in order to accelerate things beyond Moonshot and its space and power savings.
Here’s a short video that explains what The Machine’s really is and how the company came up with such a name.
The Machine combines three new principles of computing together to deliver a unique and updated vision of how computing should work. First and foremost, it combines a series of specialized cores together to help tackle different tasks as best as possible based upon whichever one does those tasks most efficiently. And considering how many different processor architectures there are out there, nobody will deny that certain ones do one thing better than another.
This is then connected with HP’s Memristor memory technology, initially in a DIMM format (but will likely morph into something entirely different) similar to current system memory. This is necessary because it offers the ability to create a universal memory architecture from top to bottom, removing storage bottlenecks and translations. The universal memory solves the next step that we take from having unified memory architectures, since those help simplify certain problems but don’t necessarily solve fundamental architectural limitations.
These processors and memristors will be connected via HP’s own photonic interconnects which promise to deliver far lower latency and better throughput than current technologies. This new architecture will also require an entirely new operating system which HP already has said they fully intend to open source in order to give everyone maximum access to it.
As Paul Teich, CTO and Senior Analyst Moor Insights & Strategy, said that The Machine, “will enable us to fundamentally rethink computer architecture at a wide range of scales.” And at this time, that’s pretty much the best analysis that anyone could give until we actually see this whole system in action, especially with the right software.
I believe that for HP, the yields of memristor based memory will be critical to their success and whether or not they can really get photonics based interconnects to scale. Most of what HP Labs has done so far is under very controlled enviroments and it would be interesting to see if the company can actually take a lot of these promising concepts and somehow get them to actually work together to accelerate computing.