Slashdot Mirror
The Microphotonics Revolution
MycoMan writes "Interesting article about photonic switching research, but there's a sentence in it that reads: 'So far, communications systems have managed to keep up because the volume of phone calls, Web pages and videostreams that optical fibers can carry is doubling every nine months, thanks in large part to the ability to squeeze more wavelengths of light into each fiber.' Doubling every nine months - is this really true?" True or not, it's an interesting article. Enjoy.
These guys aren't saying they're going to build a motherboard with fibre-optic cables. They're talking about a silicon fabrication process that is used to build waveguides (micro-fibre-optic cables, if you will) into the silicon wafer itself.
It's not the speed of light that matters here (the speed of an electrical signal is virtually identical to a light signal) - it's the switching speed. Even with the best CMOS processes out there today, there is still a finite switching time - the time it takes a transistor to go from one logic level to the other - that presents a barrier to the maximum available processing speed of the chip. With decreasing size and voltage you can improve the speed of the chip, but there's only so far you can go.
These people are exploring the likelihood that you may be able to build something analogous to a transistor that acts upon photons instead of electrons.
If they can succeed in making these feasible - then you have a technology that is potentially 1) faster than CMOS and 2) much more efficient.
That is huge. It's not just a frivolous new motherboard with lots of unwieldly wires built into it. It would be a one-piece integrated design that would in all likelihood run very cool and perhaps even faster than microelectronics ever will.
--
And I'm not saying that one is causing the other, this has nothing to do with the astrologer comment or seeing pattersn everywhere a la Pi or anything like that. But when a pattern is noticed and a possible correlation is suggested, I think it's worth asking if there is a reason for it. Maybe there isn't -- that's very possible. But maybe there is -- that's very interesting.
I'm just curious if anyone else has noticed this and put forth an explanation or refutation of the link.
DO NOT LEAVE IT IS NOT REAL
Just a sidenote, UUNET's total backbone bandwidth has been increasing by 800-1000% every year since about 1994...
In San Jose at Broadbandyear John Sidgemore (Vice Chairman, MCI WorldCom & UUNET) said in a keynote:
"Bill Gates thinks that bandwidth should be free, of course we believe software should be free."
Heiko - who works for WCOM, but is not a mindless drone
states that bandwidth increases by a factor of three every year. This means doubling every 7-8 months.
This, compared with Moore's law, has interesting consequences; among them the fact that as time goes to infinity processing power is expensive, while bandwidth is cheap. This is reflected in the differences between IPv4 and IPv6: while IPv4 has data fields tightly packed together, IPv6 spaces them out in a manner designed for easy access by software. While IPv4 optimizes bandwidth, IPv6 optimizes computational power.
Tarsnap: Online backups for the truly paranoid
I know you're more or less trolling for a laugh, but the speed of your personal net connection is totally irrelevant.
What the article is pointing out is that the total amount of traffic flowing over the web doubles every nine months. That means that if we assume everyone on the net has a 56k connection and will never have a faster one, there will be twice as many people, and hence twice the volume (assuming everyone surfs the same amount) over the internet backbones.
I found a web page, "ELECTRICITY" MISCONCEPTIONS IN TEXTBOOKS, that does a good job of explaining the difference between electric current and electric energy.
Mea navis aericumbens anguillis abundat
Interesting, but wrong. So is the other observation.
Ok we will start with 2 bandwidth and 2 processing power. Measurements dont matter, its the numbers.
For ever double the processing power makes, the bandwidth makes 2 of them. Look:
Bandwidth --- Processing Power --- Time
2 --- 2 --- 0 months
4 --- 2 --- 9 months
8 --- 4 --- 18 months
16 --- 4 --- 27 months
32 --- 8 --- 36 months
64 --- 8 --- 45 months
128 --- 16 --- 54 months
So processing power goes down 50% each 18 months compared to bandwidth.
Good luck finding any detailed information on photon based computing. You will eventually end up hitting dead links just as the information starts to get interesting. And you ask why should this be so?
You can divide electronics and computing into three generations of hardware platforms. Vacuum tubes in Zhe 30's and 40's, discreet transisters in the 50's and 60's and finally integrated circuits from the 70's to present. Each evolution in hardware platform brought a huge infusion of capital into the computing and electronics industry. Each new generation also brought about huge increases in hardware efficiency and speed.
Now what if the powers that be made optical technology available to the public? You would expect that these new computers would be much faster than silicon-wire and be very efficient.
No B fields and the ability to run perhaps tens of thousands of circuits in the same space. The paranoid might suggest that allowing the sale of such a machine would be a threat to national security, who knows. If history is any guide new hardware has often been developed and used secretly by states for a decade before the general public even hears about it. When did you first hear about the transister or the CDROM?
The propogation losses should be significantly lower... and if you can create a switching device that's efficient enough, you would be able to drive signals as far as you need (or at least as far as the "pin").
Of course, there are probably a zillion other issues that aren't occuring to me.
--
--
With 'everything' being networked and everything talking to everything else in the near future is it conceivable that our advances (such as the optical switching) won't keep up with the growth in transmitted data?
Could it reach the point where we have communication restrictions (like water restrictions :-). eg only allowed to send emails on odd days, or no emails over 3k in size :P
Recently, Corning, Inc. (mentioned in Fairley's article) announced that they will increase production at their Erwin, NY facility... by 700 jobs and $50 million. Read the press release here. Corning expects their photonics division to increase
One personal connection for me is that this plant is literally just down the street from my summer job.
What's the point of a photonic motherboard?
Well, for one thing, with a photonic based system you don't have to worry about electromagnetic interference between adjacent data tracks. That's one of the main stumbling blocks to compressing the size of circuits at the moment - at some point the individual tracks start behaving like capacitors relative to each other [that's all a capacitor is - two metal plates with a small gap between them].
With light beams on the other hand, you can even have the beams shine through each other, and it won't have any effect. That in itself would be very usefull in designing circuits. What today takes several layers could be compressed onto a single layer.
In crown glass (I don't really want to figure out the velocity in silica fibre and can't seem to find it quickly), light travels at roughly 66% of its free-space velocity. This is, indeed, very close to 2 x 10^8 m/s.
--
Not particularly... this type of stuff has happened before... Jack Kilby (TI) and Robert Noyce (Fairchild, later Intel) both applied for patents on the integrated circuit at about the same time (1959) and eventually settled their disputes by cross-licensing each other's technology.
Does TI dominate the world market in ICs today? Not really. Fairchild is still around but I don't think anyone would say they "control" the world's supply of integrated circuits.
The important thing here is that there are many different companies working in parallel on this next generation of technology. Agilent, Lucent, Nortel et al are competitors. They each want to be first to market with this stuff. And when they are first to market, they want to be entitled to collect the rewards on their considerable investment.
This is exactly why the patent system exists. To reward innovation.
So what will Agilent and Lucent do if Nortel is first? Find a rock to crawl under and die? Hell no... they'll develop their own processes.
Healthy competition is what's needed to ensure the public benefits from the technology.
I don't think the public is served in any way by refusing patents to these companies. Without patents, companies will viciously guard their secrets and forward progress is slowed considerably. With a patent in effect, others can see what one company has done and come up with novel new ideas that 1) circumvent the patent and are 2) well, novel new ideas. =)
Patents are not inherently bad. A microphotonic switch is not exactly as obvious as "one click buying". =)
--
This fortune was at the bottom of the page when I refreshed this thread:
--
Hey, this is Slashdot, nobody cares if anything is true.
Moore's law says that computing power doubles every eighteen months
The two seem to be moving at proportional rates. Interesting coincidence. Anyone wanna speculate about reasons for this? Just a glitch of the numbers, or does one have something to do with the other? How far back does this growth in communications speed go? Moore's law is claimed by some to go back in some way to the beginning of the Industrial Revolution, and with the telegraph and such I don't think it's impossible to speculate that communications has been doing the same thing.
So. Anyone care to put forth a hypothesis to explain this coincidence?
DO NOT LEAVE IT IS NOT REAL