Slashdot Mirror
All-Optical Networks: the Last Piece of the Puzzle
Esther Schindler writes "An MIT professor explains why "simple" ideas require hard science and how a gemstone might be the key to an optical network. As the story begins: 'For years, the dream of an all-optical network has lain somewhere between Star Wars and a paper cup and a string. Recent successful work on the creation of an optical diode is a virtual case study in both the physics and materials sciences challenges of trying to develop all-optical networks. It is also a significant step towards their final realization.' One answer may be... garnet. Yes, the January birthstone. 'The material that Ross and others in her field use is a synthetic, lab-grown garnet film. Similar to the natural mineral, often used as a gemstone, it is transparent in the infrared part of the spectrum. This makes synthetic garnet ideal for optical communications systems, which use the near infrared. Unlike natural garnet, it's also magnetic. ... While it works, it's too big and too labor intensive for use as a commercial integrated chip. For that, you need to grow garnet on silicon. The challenge that Ross's group overcame is that garnet doesn't grow on silicon.'"
Not the last bit to your home. That's still copper. Hence all-optical. Or something like that. This article might be too over my head to be commenting.
Brought to you by Carl's Junior.
It's amazing how much upper education is a much more ethnically-integrated environment for the finest minds in our country around the globe, but a few days ago we were told all the chinese students are spies we should watch out for (http://it.slashdot.org/story/12/04/10/0038244/fbi-says-american-universities-infiltrated-by-spies) And yet, here we are, all excited about the work some bright chinese grad student accomplished. It bugs me, that story.. a sense we are isolating ourselves from the world and what it would buy us?
CS majors know the time/space tradeoff, but they never get taught the 3rd, crucial, tradeoff of the set: comprehension!
The current paradigm is "electronic switches, optical in between." This bit of R&D is building toward a goal of "optical cables, optical switches, and only translate it at endpoints."
If the optical computer crowd succeeds, then it would be possible to have all-optical networks, to the point of only translating it to any other format for UI purposes.
As for why this is a good thing, there's a difference between the speed of electrons through copper and silicon vs. the speed of light through translucent plastics and crystals. It should also run cooler.
You also need a transistor for a logical circuit, and that's much harder.
This is, admittedly, slightly over my head as well.
I am sure that with regards to optical diodes and such, though, that they are referencing the actual hardware itself (not the dumb pipe), where switching and routing takes place. I believe the idea is that if you can build the circuits optically, they will be far more efficient -- noise/loss is far less, and the requisite voltages are much lower. No, I didn't RTFA [yet]. Get off my lawn.
Some links are optical, but at either end is hardware that converts to/from electrical signals. This happens at multiple points across the network in repeaters, switches, routers, etc. The dream of an all-optical network is one where outside of the end points you're manipulating photons the whole way with no conversions.
A cousin of mine worked for a startup back in early 2000s that made an all-optical switch, so this is remembering conversations we had back then. He wasn't working on the photonics part, though.
The enemies of Democracy are
also note that in an all optical network the analogy of a series of tubes is closer to correct than one with copper.
'...if only "Jumping to a Conclusion" was an event in the Olympics.'
As for why this is a good thing, there's a difference between the speed of electrons through copper and silicon vs. the speed of light through translucent plastics and crystals. It should also run cooler.
Ive seen varying numbers that indicate that the difference is not as large as some are supposing.
This wikipedia article seems to indicate that 82% for coax is possible, 72% for plenum cat 5.
From what I've read, and according to google calculations, the speed of light in optical fiber is about 66% of that in a vacuum. ( also supported here, and here, as a reciprocal)
In other words, copper is faster. There are certainly some benefits (removal of the interference factor, possibly running cooler, possibly lower power usage), but I dont believe signal propogation speed is one of them.
Copper is faster - which makes sense if you think about it. Light pipes require high impedance because they depend on total internal reflection to keep the light inside the pipe. The higher the impedance, the tighter you can bend the cable without signal loss.
Rule of thumb: vacuum is 3 ns/m, copper is 4 ns/m, optical cables are 5ns/m. (The long-haul cables may be a bit faster as they're made of different materials and use a somewhat different IR wavelength).
Optical-to-copper interfaces are a mess - usually the most noisey and failure-prone part of the network (aside form the end-user endpoint). All optical switching could improve reliability, power consumption, and EM noise considerably.
Socialism: a lie told by totalitarians and believed by fools.
Wasn't this the plot of Michael Crichton's book, "Congo"? The lead researcher was even a woman named Ross, IIRC...
Can you be Even More Awesome?!
It's no neodymium or iron but a magnet will still pick up quite a bit of natural garnet.
If your connection is optical all the way to your home router, then you have a fiber internet connection, not a cable internet connection.
Yes, you can do more with transistors but you can do logic with diodes:
http://en.wikipedia.org/wiki/Diode_logic
All optical switching could improve...EM noise considerably.
This ^^^ more than anything is very interesting to me. Will all optical switching and logic be inherently radiation hardened? Able to operate above the Van-Allen belts during a solar storm? I suspect the answer will be yes, until it is reduced in size to be competitive with silicon ICs, at which point it will also become vulnerable - kind of like why vacuum tubes still work with massive EMP, the tube itself is so damn big that the noise averages out.
Still, I'd bet that comparable silicon / optical systems will have different vulnerabilities and will be able to serve as complementary redundant systems.
Hmmm... a material that passes light only in one direction, and can be turned on/off using a magnetic field. Optical networks aside, I wish it was available as a coating for windows. True one-way windows!
Due to high cost, bubble memory was successful only in limited niches, so by the mid-1980s it was discontinued. Intel stopped development at the 4 Mbit level; I don't think the other vendors even pushed it that far. Late 1980s research results suggested the possibility of 64 Mbit devices. I suspect that the technology probably wouldn't have scaled much further anyhow.
More recently, IBM has been working on "racetrack memory", which works similarly to magnetic bubble memory.
What?
What the shit is a birthstone and how does this relate to optical networks?
The kind that installs and never uninstalls? We already have that.
uhm...
Surely the advantage of optical cables though is that although the speed of a signal through the cable may be slower, you can carry many more simultaneous signals through a single cable by just using different frequencies of light?
An element of delay is always introduced by translation between media. The packet coming into the building on an optical fiber hits a switch, gets transferred to copper within the switch, processed, maybe put back on another fiber to a different switch, back to copper, processed, sent on a different copper cable to a NIC, processed, transferred to the bus, and finally hits the CPU. The ultimate goal is to send a light packet from my CPU to yours without the intervening steps. I'm not really clear how many of these steps they can eliminate, but if they can just avoid changing media at the switches it would be a huge benefit.
"Think about how stupid the average person is. Now, realise that half of them are dumber than that." - George Carlin
There's nothing special about light for stacking multiple frequencies - it's really a decade or two behind what's being done with (lower-frequency) radio/microwave.
Socialism: a lie told by totalitarians and believed by fools.
Also, how the fsck are you going to send electronic mail using photons?
Escher was the first MC and Giger invented the HR department.