Beating the back-up blues
April 3, 2009That sinking feeling when your hard disk starts screeching and you haven't backed up your holiday photos is a step closer to becoming a thing of the past thanks to research into a new kind of computer memory.
Physicists at the University of Leeds and scientists at IBM Research's Zurich lab have made new advances in researching a new kind of memory, called 'racetrack' memory, which could become the standard method of storing information on home computers.
Your hard drive is a metal disc made up of millions of tiny spaces, called domains, in which all the atoms are magnetised in one direction or the other to represent binary data. Much like a record player, the disc spins around until the 'head' finds and reads the information.
Racetrack memory, a concept invented by Stuart Parkin at IBM Research's Almaden Lab, has no moving parts - instead it is the information which moves. Using a kind of physics called spin transfer, scientists use electrons (in the form of electrical current) to switch the magnetism of the domains, pushing them to a different location along a nanowire.
Recently published in Physical Review Letters, the new research holds up a magnifying glass to how tiny magnetic devices behave. Using a special electron microscope that can 'see' magnetism, scientists imaged a wall between two domains that lies in a notch in the side of the wire. This site, called a pinning centre, is where information starts and stops on its journey along the wire.
The researchers were then able to measure the current that was needed to blow the wall out of differently shaped notches.
The aim is to be able to reduce the current, and hence power, needed to move the information along the wire.
"The reason why the hard disk on your computer is likely to break is because it has moving parts which eventually wear out, but the racetrack method of storing information is much more reliable as all the parts are static," says Dr Chris Marrows, reader in condensed matter physics at the University of Leeds.
Compared with flash memory - the kind of solid state memory you find in flash drives and iPods - racetrack memory's huge advantage is on price. It is estimated that a racetrack memory in a computer would be 100 times cheaper per bit than flash.
"Magnetic racetrack memory is designed to replace the hard disk, and it's estimated that it could compete on price since it's very dense - it can store lots of bits of data on a small area of chip, as the information is stored in vertical towers," says Dr Marrows.
As well as being more reliable than hard disks, racetrack memory is also faster. There are no 'seek' times when the head has to search the disk for information, so computers would be able to boot up almost instantly.
The next stage for the team is to develop better materials from which to make the racetrack components. A fully working race track memory is anticipated to be available within 10 years.
More information: This research has been published in the April issue of Physical Review Letters.
Source: University of Leeds (news : web)
Apr 11, 2008 |
not rated yet |
0
Just because the timeline is 10 years out does NOT mean its not being looked at to put to market. With this being such a new technology, thats actually right about on par for the commercial sector. You don't seem to be including the 8 of the next 10 years the military will have this in use (aka, only 2 yrs to actual production most likely, with the military keeping it under their hats for another 8).
And there are reports of this technology going straight to transport. They're testing these trains at a gov't installation out west where the train doesn't move -- just the tracks. It's just like that guy who wanted to build a flying machine that would get really high and not have to move -- the Earth would spin by underneath him. Great stuff. I can't wait, brother!
Once again, I call it a non-starter.
A somewhat humorous non-starter that just happened to occur in the first week of April. Hmmm.
"Kryder began exploring digital storage in the 1970s as a postdoc at the California Institute of Technology. Later he spent five years at the IBM Thomas J. Watson Research Center, where he researched bubble memory, which records data by magnetizing small circles on gadolinium gallium garnet. When he joined Carnegie Mellon in 1978, Kryder continued his bubble memory work, but it became clear that the technology, used in cruise missiles and other niche applications, faced as obstacle as a mainstream product: gadolinium gallium garnet was expensive."
Bubble memory, as I recall from reading the original SciAm articles (which their search engine can't seem to find, unfortunately), was arranged pretty much *exactly* as this present article seems to describe - it consisted of tiny magnetic domains that were circulated on 'racetrack'-shaped loops, and read/written as each individual domain passed through a flux gate node. It was basically a bit-serial memory, and thus had a high latency; that, coupled with the cost (as well as the bulk of the magnetic structure surrounding it) pretty much spelled its demise.
I think this is either an 'April Fool's' joke, or some misguided attempt to regurgitate old technology by way of 'modernising' it.