Using Lasers to Speed Computer Data

Carl Bialik from WSJ writes "The start-up Lightfleet has developed an unusual way to use lasers to speed the flow of data inside a computer, hoping to break a bottleneck that can hamper machines using many microprocessors, the Wall Street Journal reports. The company plans to sell servers it predicts will be much more efficient than existing systems in tackling tough computing problems. Tasks could include automatically recognizing a face in a video image or sifting through billions of financial transactions for signs of illegal activity. These machines will attempt to sidestep some of the problems associated with parallel computation by ensuring all processors are connected, all the time."

hmm

[mastershake_phd]

Lightfleet is doing the opposite: using lenses to spread out laser beams and bounce the light off a mirror to send data around a system.

Sounds fragile, and expensive.

Re:

[fymidos]

I can feel the heat in that computer room from here ...

Re:hmm

[mikael]

The mirror will be fixed in place, so that shouldn't be a problem.

Normal interprocessor communication would require a crossbar switch or some kind of virtual network to support different grid configurations (square grid, cube mesh, torus, or hypertorus). Usually each node has a router to handle this for it. This system gets rid of the routing and just multicasts each packet of information.

Presumably receiving a data packet prevents a node from sending out another data packet at the same time. Although, this would seem to make the system act as one serial communication line. The benefits of having multiple connections is that messages can be sent between nodes in parallel. Occam had the concept of North, East, South and West connections.

Bugger

[soloport]

Moth finds nice, warm computer home. "Ooh! Pretty lights!" And a legendary term acquires new meaning.

Re:

[whargoul]

Actually, that's how we got that term in the first place http://www.jamesshuggins.com/h/tek1/first_computer _bug.htm [jamesshuggins.com]

Re:

[Sam Ritchie]

Actually, no it's not [wikipedia.org].

Re:

[Sam Ritchie]

Silly me. Link referenced by GP actually explains this. He just didn't read it (I must be new here).

The term was use during Thomas Edison's life to mean an industrial defect.

IPOL?

[jimstapleton]

IP over Laser, using a mirror as a hub? Interesting concept...

Messages from each processor, or any combination of them, are simultaneously sent to all the other microprocessors. Each receiver only picks out the messages intended for it, because of special addressing information sent with the light beams.

Yes, I know, it's not actually IP, but that's what it makes me think of.

Re:

[jd]

'nuff said [datcollaborative.org].

This solves what exactly?

[Anonymous Coward]

What - besides making your server highly susceptible to dust - does this do that HT does not?

Re:

[noidentity]

It solves the problem of how this company is going to get funding.

HyperTransport and PCIe are PtP

[jd]

This is most definitely not point-to-point.

o_0

[ukatoton]

"You don't have traffic issues and messages colliding," said Jeffrey Hewitt, an analyst at the research firm Gartner Inc.

Does anyone have any idea how they can have an all-to-all system in which messages don't collide? How is this faster than an electron based system?

Also, isn't dust in the circuits going to be much more of a concern with light based chips?

Re:o_0

[VitaminB52]

Does anyone have any idea how they can have an all-to-all system in which messages don't collide?

Build the CPU around a large free space inside the computer, in such a way that each CPU has a free line of sight to every other CPU.

Oh, and make sure no dust or insects can get inside the computer, and secure the IDE and power cables so can't hang around inside said free space.

Re:

[I_HATE_THIS]

Use different wavelength, CDMA in a beam. And I dont think line-of-sight is necessary with fiber optic cable. Sound like there is no switching, so, it is up to the local processor filter out unwanted messages (sound like Tibrv with light)

Re:

[loners]

Use different wavelengths for each sending processor.
Each processor has a single receiver that can handle all wavelengths
    or each processor has a receiver for each wavelength in the system.

You aren't suppose to connect multiple outputs to a single input. Or at least that is what I was told in my CSE course.

Seal the transmission corridors to keep dust out.

Re:

[fuliginous]

Separate send and receive channels which happen to both be able to be implemented in light which yes does interact and interfere but doesn't harm the content. Plus I it is probably possible for them all to in affect broadcast by zapping to the mirror? So everything can read everything from everywhere.

I suppose it can be quicker through 1/ travelling at the speed of light rather than electrons 2/ using a dedicated protocol that has less over heads than a generic IP one.

frequency....

[way2trivial]

think of it like this, electrons shooting down wires are like water through your plumbing.. they don't pass it two direction at once.

but, two FM stations on different frequencies can send their top forty music against each other at the same time without interfering.

Re:

[arachnoprobe]

Imagine a Beowulf Cluster of LaserSharks!

Re:

[Rastignac]

1. Install watercooling in your computer.
2. Install sharks with freakin' laser beams on their freakin' heads in the water circuit.
3. ...
4. Profit !

Back to the future

[VitaminB52]

If an insect gets inside that computer, then it can block a laserbeam. So debugging [wikipedia.org] get's back to where debugging started: keeping insects out of computer equipment, so they don't obstruct lightpaths.

Re:

[Assassin bug]

Yes, but not all insects are bugs (an enlightened term reserved for the suborder Heteroptera [bugguide.net]); therefore a more appropriate terminology would be needed. Maybe disinfest? ;o)

Re:

[glwtta]

Maybe disinfest?

Surely the opposite of "insect" is "desect"?

Re:

[instantkamera]

...or outsect?

Re:

[Duhavid]

Get the wattage up enough, and the problem will not be more than momentary.

Re:

[VitaminB52]

Yep. No more bugs :-)

Problem is, it will make the CPUs glow so hot and bright they outshine the lasers - LOL
Are you going to pay the electricity bill ? ROFLOL

Re:

[Duhavid]

Problem is, it will make the CPUs glow so hot and bright they outshine the lasers

You say that like it is a bad thing...

Are you going to pay the electricity bill

I am already tapped into your home's electrical system, so cost is not an issue.
Thanks for asking! :-)

Not really a new idea

[twiddlingbits]

Nohing new to see here..move along...the TI chip that is in the DLP TVs does this already. (from wikipedia) In DLP projectors, the image is created by microscopically small mirrors laid out in a matrix on a semiconductor chip, known as a Digital Micromirror Device (DMD). Each mirror represents one pixel in the projected image. The number of mirrors corresponds to the resolution of the projected image. 800x600, 1024x768, 1280x720, and 1920x1080 (HDTV) matrices are some common DMD sizes. These mirrors can be repositioned rapidly to reflect light either through the lens or on to a heatsink (called a light dump in Barco terminology). The rapid repositioning of the mirrors (essentially switching between 'on' and 'off') allows the DMD to vary the intensity of the light being reflected out through the lens, creating shades of grey in addition to white (mirror in 'on' position) and black (mirror in 'off' position). This system is just a new application of a technology invented in the late 1980's. No reason it's groundbreaking and no reason it shouldn't work in theory. Solid state lasers are very reliable and have a long life time. However, I don't know of any chips that have the ability to directly receive laser light pulses from a source and convert them to 1's and 0's. And keeping the lasers, mirrors and receivers aligned might be tough.

Re:

[mark3748]

And keeping the lasers, mirrors and receivers aligned might be tough.

The answer is fiber, of course. Same thing with the dust and other obstructions

Re:

[twiddlingbits]

I thought this design did away with the fiber connections, that was the benefit.

Re:

[mark3748]

it does, but that's not exactly the most efficient way to do it. I wasn't saying that's how they are going to do it, I'm just saying that's how to solve the problem.

The light is going to travel the same speed through fiber as is would through air, so I have no idea why they would want to get rid of fiber connections.

The only realistic way you could do this outside of a laboratory environment without fiber, is transferring information between parts of a single chip, such as multiple core processors and t

Re:

[twiddlingbits]

Not exactly, it would have to be a sealed unit with shock/vibration isolation. There are similar devices such as a ring-laser gyro that transmit laser pulse w/o fiber. It all adds up to being a nice lab toy, it's far too expensive and impractical for the desktop or server environment.

Re:

[Dread Pirate Skippy]

TFA specifically states they won't be using fiber, and they will be using mirrors. So fiber might be the answer, but probably not to the question at hand.

Re:

[twiddlingbits]

from TFA to prove my point.... The design is particularly efficient at sending "all-to-all" messages between chips in a system, said Bill Dress, a Lightfleet senior scientist and co-inventor of the technology. Because the system sends light through air, Lightfleet avoids the need for wiring and associated switching circuitry and software, he adds.

Re:

[Trogre]

However, I don't know of any chips that have the ability to directly receive laser light pulses from a source and convert them to 1's and 0's.

Uh, wouldn't that be a phototransistor? How do you think fibre optic channels work?

Re:

[twiddlingbits]

The context of the article was CPUs, so that was my meaning when I said "chip". Next time I'll say CPU CHIPS. Name me a mainstream CPU with a phototransistor on-board.

Re:

[Pooua]

twiddlingbits: "Name me a mainstream CPU with a phototransistor on-board."

Who said it had to be a mainstream CPU chip? Who said the phototransistor had to be on the CPU chip itself? The article is vague on such details.

"Lightfleet said new servers it plans to begin selling next summer, which will use 32 dual-processor chips from Intel Corp., can do such feats easily....

"... Each microprocessor is installed with a laser transmitter and a set of devices that receive beams of light..."

Now, it is already well-k

Re:

[twiddlingbits]

A DLP actually does encode, it's either 1 (light on a area) or 0 (light off an area), that is a simple code. DLPs are fast, but certainly not in the GB/Sec range such as a FibreChannel type connection. But those fast connections take special (expensive) hardware.

How will each CPU seperate the data meant for chip X from that meant from Chip Y since it seems to be to be a broadcast bus not a point-to-point connection.(i.e. there has to be a protocol)? Is the decoder on-board the CPU (extra silicon) o

I know this one!

[Neoncat]

Hey, I have played this game on c64. I think it was Deflektor.

Hate to be a negative nancy

[bperkins]

I'm a bit surprised the Wall Street Journal would more or less paraphrase a vacuous press release and pass it off as an article.

I'm less surprised (but still surprised) that slashdot would pick up such a piece.

My suggestion for a tag:
pressreleaseaseasjournalism

Re:

[sensei moreh]

My suggestion for a tag:
pressreleaseaseasjournalism

+1

Re:

[EveLibertine]

You both missed out on the typo:
pressreleaseaseasjournalism = press release ase as journalism

Whatever the hell that means.

You think so?

[jd]

I'm restricted by NDA and by corporate policy in what I can say, so I won't comment on the article itself. I will say though that most newspapers are nothing more than paid advertising that run a few real articles as typing exercises.

Passive Star Networking

[PSaltyDS]

The US Navy used to do networking over a Codenoll Passive Star network [dbasinc.com]. The modified 10Base-FL NICs sent transmit pulses to passive hub, which optically coupled all the rcv/xmt ports together in what was essentially a fused glass blob. Codenoll calls it 10Base-FP.

The useful thing about it was being completely unpowered. The passive hub could stuffed into/behind anything where the fiber could reach it and there was no configuration, power, management, etc. Of course, those were also its weakness: no configuration, management, etc. A lot of these were installed in the early 90's, but I don't think the Navy uses them any more.

Re:

[moosesocks]

Verizon uses a similar technology for its FTTP network. Each node operates on a slightly different wavelength, which is separated by a passive prism-type device at the street level. Effectively, this means that in areas where FTTP is deployed, there will be no active electronics on the poles, which has numerous advantages.

Of course, speeds are *much* faster than 10mbps.

Duh.

[LoudMusic]

Like, duh, everything is faster with freaking laser beams.

Thanks

[UbuntuDupe]

I just wanted to give a big thanks to taggers for not putting the "sharks" tag on a laser-related article. At least, not yet anyway.

Re:

[mobby_6kl]

I wasn't going to, but now that you said it, I've tagged this article "sharks".

I don't see what you problem is with this tag anyway. A few years from now you'll be able to search for "sharks" to get all laser related stories.

Re:

[Lorkki]

And what about shark-related stories?

Two words

[gr8_phk]

Snake Oil. Actually, they don't even have a product yet. What's the term for raising money for an idea that will never fly?

Re:

[porpnorber]

The DMCA already provides legal protection for snake oil in the form of copy protection schemes. Even if this technology doesn't work in the least, all it would take is for a Digital Millennium Networking Act to be passed saying that nobody is allowed to demonstrate flaws in a networking technology.... When there's a government to supply the truth, who needs a working product anyway?

more info from patent app

[shizzle]

If you're interested in the details, check out US patent application 20040156640 here [uspto.gov].

It's brutally long (51 pages!) but provides a lot more details.

Basically each node has N (or N-1) receivers spaced slightly apart, along with a single transmitter. These receivers & transmitters are all in the same plane opposite a mirror. Every node can transmit simultaneously. The different angles at which the transmitters hit the mirror cause the beams to focus on a different receiver within each node's arra

Faster computers only used to Protect Public?

[mathx]

Why are both examples of how a faster computer could benefit society framed in a 'law and order' and 'keep the public safe' context? Smart marketing that plays upon the already nicely laid groundwork of the American administration?

I wonder if faster computers could help the average person in any way at all.. cant think of any examples tho. :/

-math

Hmmm.

[jd]

You're lucke that you posted that anonymously. :)

Re:

[jd]

My mistake. That really was an anonymous howard.

Actually, that's pretty cool...

[Tarlus]

It will literally be like computing at the speed of light!

In using light to relay computer data, the only barrier that would be left would be to reduce the distance between two communicating nodes (or reduce overall size of the technology).

Taking something out of Networking...

[mlw4428]

Why not use fiber cables to connect the processors? It eliminates the problem with dust as well as the need for mirrors that take up system space.

a "bus"?

[chanceH]

so, I actually skimmed the article.

I'm way to lazy to actually read all of the words in it.

But isn't this just a "bus"? but its done with lasers?

I'm not really seeing the advatange, except that you get to use the word "laser" in your IPO?

Re:

[Fission86]

Yes, it is a bus, but it transfers at the speed of light

Re:

[jd]

Find me a multicast-capable bus that electrically scales and I'll find you an answer.

Tbps Cell IO

[Doc Ruby]

The Sony/IBM/Toshiba Cell uP already has an onchip token ring running at 204GBps [wikipedia.org]. Sony is reportedly developing an optical interconnect to join devices together, presumably at that speed (1.64Tbps).

What interconnects already exist anywhere near that speed? 10Gbps ethernet is about $400 per card on a PCIe bus, or 2x10Gbps on a card for $700. Is there 100Gbps for sale today at any price? Any other >10Gbps signalling on a PCIe card, or even on a motherboard?

doomed!

[casehardened]

Not to be too harsh, but: 1) Nobody on their management, board of directors, or technical board has an optics background 2) They're doing this in free space. That's fine in a _mechanically stable environment_ 3) Computers are not mechanically stable enviroments. PCB boards flex. Things heat up and cool down. Everything moves around over time 4) Focusing a received laser beam to a detector requires precise alignment. A 1 GHz detector is ~ 0.4 mm square. A 12 GHz detector is 25 microns square (New Focus Optic

Hurry up ...

[PPH]

... and finish coding that module or I will shoot you with my laser.

Waddya know? It works!