Slashdot Mirror
Photonic Switching to Boost Internet Speeds
Da Massive writes "Researchers at the University of Sydney have developed technology that could boost the throughput of existing networks 100-fold without costing the consumer any more, and it's all thanks to a scratch on a piece of glass.
After four years of development, University of Sydney scientists say the Internet is set to become, on average, 60 times faster than existing networks.
According to the Centre for Ultra-high bandwidth Devices for Optical Systems (CUDOS) at the University's School of Physics, the scratch will mean almost instantaneous, error-free and unlimited access to the Internet anywhere in the world."
That's all very well and good, but the last mile over here is over copper and based on the inaction of the TelCo, and the lack of REAL competition, will remain copper for another 100 years. So no matter how fast the IP packet takes to get to the exchange, it will be slowed down.
Take Nobody's Word For It.
I thought it was about photonic switching, not the actual fiber cables itself. Basically there's loads of dark fiber because the switches aren't fast enough or powerful enough to use it all. A photonic switch can make use of it all, and also make use of the full capacity of the fiber rather than have the line speed limited by the switches.
So, you'd need new switches at either end of each fiber cable. I don't know how often backbone switches are replaced, but I could see that happening within 3 years.
This is great for the rest of the world where the technology will be implemented. Here in the States, the mega-elite who stand to lose billions if they lose control of the throttled internet will suppress this somehow. America, the most powerful throttled (health care and internet) country in the world.
A conventional electronics packet switch is a store-and-forward device. It receives (at least the header parts of) a given packet, stores the data, decodes it, decides what outgoing ports the packet needs to be re-transmitted upon, composes a new header part, and re-transmits the entire packet on the outgoing port. This means that the packet itself must be buffered, and there is necessarily an overhead latency of many bits (at least the length of the header of the packet) between the input bitstream and the output bitstream.
In an optical switch, the optical data is split, so that a duplicate optical pattern goes down two paths simultaneously. One path is basically a many-turns coil of optical fibre, so that it will take a few picoseconds for the carrier-light to transit the length of this coil.
The other optical path goes immediately into a detector and optical logic switcher (if I may coin a new term, "optonics", if you will indulge me), so that the header information is decoded and an optical switch is set to the correct output leg, and a new header is composed and transmitted, just in time for the slightly-delayed carrier-light of the main bulk of the packet to arrive from out of the coiled length and be appended to the outgoing header.
The technology requires fine-tuning of the length of the coil of optical fibre to match the switching latency of the header/decoder/re-generator part.
The entire latency of the packet's transit through the entire optical switch is of the order of one sixtieth of the latency of the highest-performance conventional electronics switches.
Neat innovative new technology brought to you from Oz. Now that is really going to surprise a few arrogant yanks when they eventually figure it out, is it not?
Article contains link to CUDOS and the description of the research (not a scientific paper yet) is pretty easy to reach.
http://www.cudos.org.au/cudos/research/Chalcogenide.php
How do they do it is not straightforward and IMHO impossible to explain in popular science article.
This has been around since at least October 2005. A slightly better article that contains a little more information (albeit its still kinda vauge) is here
Most of the core devices in carrier grade networks have an expected five year lifetime before being moved closer to the edge. May system have edge devices that were bought 10+ years ago.
Without many more details it would be impossible to judge when this would be avaialble to the average consumer.
That is not to say that you might not see an performance boost because google was able to upgrade the youtube connection.
Since the title says 'switching', I'm guessing it's routing efficiency. I've not RTFA but I'm planning to in a minute if that makes things any better ;)
Totally agree with you about the latency thing. I was reading reviews of Battlefield Bad Company for the PS3 the other day, someone has said "this game has no lag!". While the server/client communication may be more efficient than other similar online games, there's no way it will have 'no' lag (which I would equate to latency). And in fact, when I've been playing it myself some games have had some serious lag anyway, which I doubt was my own connection as connecting to a new server tends to sort out the problem. I seem to get a lot of dropped packets on some servers as my character actually drifts backwards :s
Most people just don't know what they're talking about when it comes to things like networking, but they'll try to pretend they do*. What else is new..
* that's what I do anyway \o/ these suckers are paying me to administrate their network and I have no idea what a subnet mask looks like, or why a subnet would even want to hide its identity
which is totally what she said
I am guessing that it is all still a bit secrety, but basically the technology will allow optical network switches instead of electronical.
Optical circuits.
"The scratched glass we've developed is actually a photonic integrated circuit," Eggleton said.
"This circuit uses the 'scratch' as a guide or a switching path for information - like when trains are switched from one track to another - except this switch takes only one picosecond to change tracks. This means that in one second the switch is turning on and off about one million times. We are talking about photonic technology that has terabit per second capacity."
An initial demonstration proved it possible to achieve speeds 60 times faster than existing local networks.
Mit der Dummheit kämpfen Götter selbst vergebens
The paper showed up. Apparently it makes use of the Kerr effect. I'm speculating that it's more specifically kerr-lens modelocking.
One of our competitors trademarked the term "hypothesis". From now on, we will call them "boneheaded ideas".
five meters on 238 litres of petrol
Where in the world did you come up with that conversion?
40 (rods per hogshead) = 0.843539102 meters per liter
So you could go slightly over 200 meters with 238 liters of gasoline. Sheesh, you were off by a factor of 40!
Ah, I guess you calculated 1 rod per hogshead.
Any sufficiently simple magic can be passed off as mere advanced technology.
Here's the technical explanation: http://www.cudos.org.au/cudos/research/Research.php ...and here's another topical piece:
http://news.ninemsn.com.au/article.aspx?id=594743
Yep, I'm in the research group concerned. It was pretty funny to look at the ways in which this story was distorted and Eggleton's soundbites were distorted.
He was talking about two things:
a) This new advance, which is about being able to pull a 10Gbps signal out of 64 time-division multiplexed 10Gbps signals (and was actually done by a group in Denmark/China, with one of our waveguides.
b) The Photonic Chip, which is our long-term goal, of all-optical signal processing (e.g. an optical router).