back to article All-optical RAM to clear comms bottleneck

Japanese researchers are claiming a breakthrough in all-optical memory, one of the key bottlenecks remaining in the optical communications world. The high throughput of optical communications systems brings its own problem: any function that can’t be performed in the optical domain demands an opto-electric conversion, creating …

COMMENTS

This topic is closed for new posts.
  1. Andrew Alcock
    Holmes

    Oh dear, oh dear - you don't remember your history do you

    The world's second computer - the Manchester Mark 1 (disclosure, I went to Cambridge, and so would argue vociferously that Cambridge beat Manchester in the race) - had the "William's tube" CRT memory in 1946:

    http://www.cedmagic.com/history/williams-tube.html

    And here is another photograph of CRT memory:

    http://en.wikipedia.org/wiki/File:CRT_memory.jpg

    1. Captain Save-a-ho
      Facepalm

      Re: Oh dear, oh dear - you don't remember your history do you

      CRT memory wasn't optical. RTFM.

      1. Anonymous Coward
        Anonymous Coward

        Re: Re: Oh dear, oh dear - you don't remember your history do you

        isn't a photograph optical memory - for very broad definitions of memory.

        1. atippey

          Re: Re: Re: Oh dear, oh dear - you don't remember your history do you

          also, its definitely write only

          1. kiwimuso
            Joke

            Re: Re: Re: Re:Oh dear, oh dear - you don't remember your history do you

            Write only?

            Does the memory suffer from Alzheimer's then?

            Can't see the point of storing something in memory if you're not going to access it again.

        2. Andrew Alcock

          Re: Re: Re: Oh dear, oh dear - you don't remember your history do you

          http://en.wikipedia.org/wiki/File:Chauvet_cave,_paintings.JPG

          Given the right conditions, painting memory can last 30,000 years, a rather better MTBF than magnetic media.

      2. Andrew Alcock

        Re: Re: Oh dear, oh dear - you don't remember your history do you

        Well, then, neither was the Mellon optical memory linked to by the author:

        "To form a complete memory the system was arranged to be regenerative, with the output of the photocells being amplified and sent back into the CRT to refresh the cells periodically."

        The difference is simply that the decay on the Mellon system was longer than the decay on the Williams CRT, so the refresh rate was lower.

        My favorite highly impractical early computer memory was http://en.wikipedia.org/wiki/Delay_line_memory#Mercury_delay_lines

        1. John Smith 19 Gold badge
          Boffin

          Re: Re: Re: Oh dear, oh dear - you don't remember your history do you

          "My favorite highly impractical early computer memory was http://en.wikipedia.org/wiki/Delay_line_memory#Mercury_delay_lines"

          *Awkward* certainly (and no doubt in breach of numerous H&S regulations today).

          Impractical. No. Multiple systems used them (IIRC the LEO commercially sold mainframe)

          You're just so used to everything being in a chip you don't realize how *diverse* the history of memory systems is up until at *least* the late 1960's. Multiple mfg fought it out with various concepts for both main and register memory.

    2. Andrew Alcock
      Facepalm

      Re: Oh dear, oh dear - you don't remember your history do you

      D'oh - should have said 1948, not 1946.

  2. Andy 70

    hmmm, so would pron be stored as pron in optical memory? would it even need a display device?

    are we going to see a researgence in Boots faux red/green/blue leather photo albums as "optical storage media" for fambily fauxtografy?

    has grannie with her years old dusty shelves and shelves of the afformentioned picture storage devices, got the drop on us all?

    the mind boggles with the imaginary possibilities.

  3. Jon Double Nice

    'a material based on an indium gallium arsenide strip buried in gallium arsenide'

    Oh come off it, that one is sooo obvious...

    My favourite is this one: http://en.wikipedia.org/wiki/Delay_line_memory#Optical_delay_line_memory - just bounce some lasers off the moon!

  4. Ken Hagan Gold badge

    30 nanowatts you say?

    Since the article also mentions a refresh time of a microsecond or so, I assume that's the ongoing cost of maintaining 1 bit of data with this new scheme. That's several hundred watts per gigabyte and a few kilowatts for a half-decent server. (Come back Pentium 4! All is forgiven!)

    I know this is research and things will improve. I just though someone ought to flag up that we are several orders of magnitude short of a usable system.

    1. Graham Bartlett

      Re: 30 nanowatts you say?

      Doesn't really stop it being useable, it just puts an upper limit on the amount of RAM you can have if you've only got a domestic mains connection. Personally I think the more important "orders of magnitude short of a useable system" element is that they've only got four bits of optical RAM at the moment...

    2. Bob H
      Facepalm

      Re: 30 nanowatts you say?

      The primary role for such a technology isn't really computing but in switching. Network switches use small quantities of RAM as buffers on the switching fabric. Without buffers on switching fabric you need a blocking infrastructure or perhaps a TDMA mechanism with an RTS signal back to the source.

      If we have pure optical switching with pure optical buffers then we can have pure optical core switch fabric and provide much faster infrastructure in WANs, MANs and large offices. This is really mostly carrier grade technology, which is also why NTT is so interested.

      The idea of an optical computer might be of interest to us mortals, but in the end it will be the carriers who pick this stuff up first. Much like the early computing and networking technology.

  5. Fred Flintstone Gold badge

    Uh oh..

    I recommend exposing this memory to a bunch of 5 year olds with torches before this goes anywhere near production :-).

  6. John Smith 19 Gold badge
    Coat

    A non-electric optical RAM.

    From time to time I've wondered about a completely non-electrical computer system.

    We know a mechanical difference engine *can* be built and does work.

    But I've always found a non-electric VDU a *real* challenge.

    Even restricting it to a single font/no graphics device you have to compose a *dynamic* image and allow its editing and re-display.

    This is a start but it's a long way to go.

    Mine will be the one with a copy of Charles Stross's "The Fuller Memorandum" in it.

    1. The First Dave
      Boffin

      Re: A non-electric optical RAM.

      Don't you remember the bid display panels that they used in Railway stations a couple of decades ago? Sure, they were electrically driven, but the actual display was a purely mechanical open/closed "flip" display. A purely mechanical version would make the difference engine look simple, but it is entirely possible.

      1. John Smith 19 Gold badge
        Happy

        Re: Re: A non-electric optical RAM.

        "Don't you remember the bid display panels that they used in Railway stations a couple of decades ago? "

        Fair point. Can't remember if they could handle the full alphabet or it it was just numbers.

        IIRC something like it is also used in Gibson & Stirling's "The Difference Engine" with its IT revolution happening in Victorian Manchester, whose developers are called "clackers" as a result.

        I'd been thinking of something more opto-mechanical like Vanevar Bushs Memex proposal.

  7. John Smith 19 Gold badge
    Boffin

    Mellon memory. Time for a comeback?

    http://en.wikipedia.org/wiki/Mellon_optical_memory

    With the *greatly* improved fabrication methods and materials tech of today this idea *might* be substantially more viable today.

    Note that at *no* point does it have a need for a single crystal material. Everything *could* be done using the larg area techniques developed for flat screen TV's.

    Just a thought.

  8. Alan Bourke

    Bubble memory

    FTW.

  9. Anonymous Coward
    Anonymous Coward

    "... the optical cavity approach consumes ... 30 nW"

    Thanks a lot for undoing years of scientific training by seeding my mind with the El Reg Official Units. Am I the only one whose first reaction to the statement of the title was "wait a minute, what are you doing, using a unit of surface area for reporting a measure of power??"?

    1. Tom_

      Re: "... the optical cavity approach consumes ... 30 nW"

      30nW is more or less 186 square fathoms.

      Happy now?

  10. Anonymous Coward
    Go

    The Bleeding Obvious Optical Memory

    ...would be using loops of fiber. One meter will easily store 30 bits (10 Gbit/s signal rate - 3 cms per 0,1ns) per wavelength. So 100 meter would store already 30kBytes, which is already quite a bit for a high-speed switch. I don't exactly know how signal regeneration would be done, though.

    Taps could provide a limited form of addressing.

    1. Christian Berger

      Re: The Bleeding Obvious Optical Memory

      Well sure yes, but you need to amplify light every couple of 10th of kilometers. You can do that optically, but not to often, since those optical amplifiers have a lot of noise. So in the end you'd have to regenerate the signal by converting it to electronic signals and then back to light... that way you loose the advantage of being all optical.

      Of course photonic networks are still a pipe dream and I'm currently skeptical how useful those will be.

  11. John Smith 19 Gold badge
    Boffin

    Oh an optical delay line

    Which is the light analog to the ultrasonic delay line using Mercury or lump of glass in PAL TV sets (that gave 64micro secs at a TV bandwidth of about 5.5Mhz).

    An optical delay line is about 3.3ns per metre. So that microsecond memory would be about 300m long. But you'll need a *lot* of them for parallel access (IE images).

This topic is closed for new posts.

Other stories you might like