Subscribe / Unsubscribe Enewsletters | Login | Register

Pencil Banner

Seeking the Fourth Circuit Element

F.Y. Teng | May 25, 2010
A conversation with the czar of R&D at HP Labs about taking memristors from paper to production to mobile computers to enterprise infrastructures.

That becomes a major choke point in computation, especially the modern types of computation that go on with search and Cloud and other types of issues.

So the first area that we see memristors playing a role in enterprise computing is as an intermediary, or as a buffer if you will, between DRAM memory and hard disks. A lot of people are looking at solid state storage and the main candidate for that right now is flash memory. But flash poses several problems, its actually quite expensive and it doesnt last very long. It wears out very quickly, so its not the greatest candidate for this buffer although people are using it because they have to have something. They need something to sit there. So the first thing we see is that flash can be a solid state memory that resides between DRAM and spinning hard disks and gives us an ability to store large amounts of data in a non-volatile cache that we can exchange very rapidly between the memristors and the DRAM in order to speed up the total computation cycle.

That is effectively getting the foot in the door. But as time goes on, we believe that this technology will become more reliable, faster and less expensive. Andbearing in mind the Clayton Christensen idea of the innovators dilemmawe essentially see that we can be replacing more and more of both DRAM and hard disks with memristors-based storage since memristors storage is both random access like DRAM and also has time frames access times and speeds that are similar to DRAM. But at the same time it is also permanent storage, its non-volatile.

Eventually, we think its possible that memristors can take over from DRAM and hard disks and provide a single unified data storage platform, what some people may call universal memory, making it possible and easy to flatten out the data hierarchy, and essentially make every bit of information people want to know available in a random access fashion, therefore making it very fast and speeding up overall computation.

Please fill in the blanks between when your labs decided to pursue memristor research and development and whats been happening since?
Weve been working on this for close to 13 years, so we have been anticipating this issue of the gap widening between DRAM and hard disk storage, and also other problems with memory and storage, for a very long time. And since HP has been really one of the largest consumers of DRAM and storage for a long time, this was a significant issue for us internally. Thats why we were doing research in this area in the first place. And so it was in 2008 that we finally understood this class of what some people call ReRAM, Resistance Excess Random Memory. Other people have called it many different things, bipolar switching for one. This is something thats been studied since the 1960s and people just didnt understand that this was memristors because they didnt understand the physics of how these things operated.

 

Previous Page  1  2  3  4  5  6  Next Page 

Sign up for Computerworld eNewsletters.