Closing the Gap To Improve the Capacity of Existing Fiber Optic Networks

cylonlover writes "A team of researchers working through Australia's Centre for Ultrahigh Bandwidth Devices for Optical Systems (CUDOS) has developed data encoding technology that increases the efficiency of existing fiber optic cable networks. The researchers claim their invention, which packs the data channels closer together, increases the data capacity of optical networks to the point that all of the world's internet traffic could be transmitted via a single fiber."

Comments

High Speed for who?

[trazom28]

Now we just need more locations to actually *have* fiber, or some similar high speed bandwidth. My in-laws can only get celluar (unreliable), and satellite isn't worth it. They are just within range of DSL if the phone company would do the upgrade - and there are several customers on their street that would happily switch - apparently not enough for them to spend the dollars to do the upgrade.

Upgrades only are cool if everyone has the opportunity to use it.

Re:High Speed for who?

[The Wild Norseman]

This is for core long haul transport, not your in-laws house.

My in-laws' house is for core long haul transport, you insensitive clod!

Re:High Speed for who?

[medv4380]

Except fiber happens to be the preferred medium for the High Traffic parts of that Highway. Reducing congestion in the higher trafficked areas improves the performance of the network as a whole which improves your in-laws network performance as well. Try not to look a gift horse in the mouth. I'm sure the phone company loves not being able to turn on any of that lovely fiber they over provisioned in the 90s because of these advances.

Re:High Speed for who?

[Miamicanes]

> My point being, why upgrade the long haul transport when we can't get everyone on?

Because dark fiber isn't equally-available to network service providers. Lots of small, disruptive companies managed to buy up a couple of dark fibers in the days following Worldcom's collapse, before bigger companies snapped up most of the remainder to hoard and maintain scarcity. If a dozen small, disruptive companies that own a fiber or two apiece can make that one fiber do the work of 8, it really doesn't *matter* whether it would be cheaper to just use cheaper and simpler modulation methods on 8 dark fibers than to use exotic equipment to multiplex 8 times the data onto one. In the real world, if you aren't Verizon or AT&T, the cost of acquiring 7 more fibers is likely to be a lot higher than the cost of buying expensive electronics gear and stacking 8 times the data onto the one you already have.

Here's another example: back in the late 70s, the amount of money a big company with offices in New York and Chicago paid to MCI for a virtual trunk line connecting their PBX systems in the two cities was WAY more than what it cost AT&T to actually own and operate a comparable inercity trunk line.. but the amount charged by MCI was less than AT&T charged, and it ended up being several orders of magnitude cheaper for employees in New York to make Chicago calls by picking up their desk phone, connecting directly (via MCI) to their company's PBX in Chicago, hitting 9, and dialing the local Chicago number, than it would have been to have just directly placed a long-distance call through AT&T and paid their per-minute charges and taxes to make the call.

As my dad explained it to me (he used to work for MCI), it was technically against AT&T's TOS back then to run your own intercity bridge and use a PBX in one city to make calls from another... but the Carterfone decision made AT&T's authority to dictate such terms legally questionable, the FCC was in no mood to enforce such terms anyway, by the time the feds started to care about lost excise tax revenue even medium-sized companies were doing it, and AT&T was hoping that if it quietly behaved itself and didn't cause a fuss, it might be able to avoid getting broken up. Later, MCI built switching centers where they allowed companies like IBM and Ford to just lease a colocated PBX (maintained by MCI) so they could purchase leased trunk lines into cities like Miami where they didn't have a direct presence, followed by a whole chain of incremental steps that allowed companies to share their local POP, trunk lines, and pool of local POTS lines with other companies, until finally MCI just started offering outright bulk prepaid long-distance service to companies. At that point, you still had to jump through hoops that basically boiled down to "dial a local number to connect to a local PBX, dial the desired number, let the system switch you over via private trunk lines to the destination city's PBX, which connected you to a local phone line, dialed the local number on your behalf, and connected you to it"... but it worked, and allowed large (and eventually, medium-sized, gradually extending to smaller) companies to place long-distance calls for a fraction of what AT&T charged.

Put another way, the way disruptive companies like MCI did an end run around AT&T was more expensive than the internal efficiencies enjoyed by AT&T, but ended up being cheaper for end users because AT&T didn't pass those efficiencies along, and instead milked them for every mononopolistic rent-seeking penny they could. The same is true with fiber. If you own a fiber and can use it however you please, being able to multiply its capacity is HUGELY disruptive to larger companies whose business plan is to maintain artificial scarcity and keep prices high.

More importantly, much of the local impetus comes from disruptive companies like Google who have more backhaul than they know what to do with who then turn around and make it available to end users in a cit

Putting more D in DWDM

[Shatrat]

Looks like a Super Channel implementation. Not really a novel concept for next gen > 100Gigabits per channel DWDM systems.See here for example. http://www.lightwaveonline.com/articles/print/volume-29/issue-2/features/superchannels-to-the-rescue.html
More power to them if they're making good progress, though.

Latency?

[mortonda]

all of the world's internet traffic could be transmitted via a single fiber.

Um, sure, that's easy.... but how long will it take?

Re:Latency?

[gstoddart]

Whoops, speed of light / refractive index, otherwise time travel possible.

Time travel is possible -- it's just that so far we've only figured out the forward, linear kind. ;-)

Re:Latency?

[Shatrat]

Or, the non-linear time dilation effects of tequila.

Re:Latency?

[jc42]

all of the world's internet traffic could be transmitted via a single fiber.

Um, sure, that's easy.... but how long will it take?

And how long would it take for the first backhoe to come along?

Re:Latency?

[Anonymous Coward]

Tequila, the only alcohol you have to be drunk to drink.

Re:Latency?

[jc42]

[H]ow long would it take for the first backhoe to come along?

Funny, perhaps, but there's a serious problem hidden in the phrasing "all of the world's internet traffic could be transmitted via a single fiber." Most people would naturally consider this a major achievement, but in fact it would be a major mistake.

Back in 1986 (12 Dec, I googled it ;-) all of New England (the one in the US, not Australia ;-) was cut off from the rest of the network for half a day. This was thought unlikely, because there were seven different cables connecting the northeast to the rest of the country, and what was the probability that they'd all be down at once? But the comm company that supplied the "cables" (AT&T) had implemented "virtual cables". They had cleverly routed all the traffic through a single cable, and a worker in New Jersey cut it.

This has become a textbook example for a lot of design issues. The main one is that the software has to be able to see down into the lower "levels", to make sure this hasn't been done. If the high-level software can't see down into the physical layer, as the "design" folks often insist is the only correct design, then there is no way to write code that detects such problems and throws out warnings that a failure is likely.

Another lesson is that you can't trust the hardware. The people responsible for the hardware will do this sort of thing whenever they can get away with it. It saves them money, and they don't even have to lower prices for customers who can't detect what's been done.

The Internet was designed from the start with multiple routes everywhere. The commercial world hasn't implemented this part of it very widely. As a result, there are often only a single path between two sites, and the network is constantly plagued by outages that wouldn't happen if the software could just "route around the damage", and the saying goes. But hardware is expensive, and companies install the minimum that they can get away with.

The world is slowly coming to rely on the Internet. If we want it as reliable as its designers (the US DoD) intended, we should object to every suggestion that any traffic be handled by a "single fiber". Yeah, it might sound impressive, but it's actually dangerous. Instead, we need to find ways of forcing the hardware companies at the bottom to build the needed physical redundancy, so that when that single fiber fails, the software will reroute the traffic in a millisecond or so, and only the company's workmen will have to know that someone cut it with the proverbial backhoe.

Re:Yeah

[mjwx]

Mr president, we cannot afford to have a fibre gap.