Terabit Ethernet Inches Closer To Reality

An anonymous reader writes "Researchers from Australia, Denmark, and China have combined efforts to show the feasibility of terabit-per-second Ethernet over fiber-optic cables. The solution involves a photonic chip that uses laser light for switching signals, and a form of the exotic material type, chalcogenide, or arsenic trisulfide."

Comments

That's an aweful lot of porn.

[tjstork]

Not that I would ever use a terabit connection for porn... but uh, when's that coming out again?

Re:That's an aweful lot of porn.

[Wrexs0ul]

Not really. It's a well-known fact a lot of innovation is driven by the porn industry. This stuff is probably being sponsored by the Ultraporn coalition to put their digital media online.

Imagine streaming video so clear you can actually sense the actresses' emotional issues!

-Matt

Re:That's an aweful lot of porn.

[Anonymous Coward]

Being someone who works at a porn company with multiple dedicated lines buried under the ocean, I can say this is very true. We test all the equipment we have to the limits.

I worked for a lot of mom and pop companies that thought they had problems.

We are pretty much a dedicated Foundry and Cisco debugging team.

When a single server gets over 10,000 hits a second (yes, second, not minute) - it tends to stress your equipment.

Times that by a few hundred servers and you get the idea.

I used to deal with simple PHP and Apache issues before. Performance? Was never an issue.

Now half our stuff is written in heavily optimized C, our kernels are heavily tweaked and even Squid isn't fast enough to keep up.

We even have our own custom caching software.

Re:That's an aweful lot of porn.

[MBGMorden]

On a large enough sample set, correlation implies a relation other than chance, and thus should be investigated. Otherwise you can keep screaming "Correlation is not causation" at every piece of data every produced and try to claim that we can never claim results.

After all, if I state that "Each time a plant is deprived of water and sunlight it dies.", stating "Correlation is not causation" is complete nonsense. We've observed over a large enough sample set that yes, in this case correlation damn well IS causation. Effectively, your only argument here should be whether or not the sample size is large enough.

Re:That's an aweful lot of porn.

[RiotingPacifist]

Prostitution, recommended by 9/10 doctors as less virus ridden than using IE for browsing porn.

Re:That's an aweful lot of porn.

[drinkypoo]

The porn industry gets it more then the main stream movie industry does

I would imagine.

as far as getting your entrainment out their.

Out their what? Wait, maybe I don't want to know.

Re:That's an aweful lot of porn.

[mmontour]

That said, as it become more mainstream it will draw people less screwed up. What the industry really needs is a union.

Bad idea - you'd end up with all of the acting jobs going to the women with the most seniority.

Re:no good

[BadAnalogyGuy]

I only need inches, if you get my drift.

sweet

[_avs_007]

Now I can finally get started on building my holodeck.

What value?

[pig-power]

Tera ethernet... 5-25 gig monthly caps... "I used my monthly cap in 31.65 seconds..UH O..."

Re:What value?

[Shakrai]

You also have to account for your neighbor who is addicted to porn and downloads it constantly seeding at 100% for days on end.

Hey, don't talk about me like that when I'm not around ;)

Re:What value?

[Mister Whirly]

Well, it would only take .2 seconds if the sending server was serving you up packets that fast. Somehow I doubt you would get maximum throughput speed. I know I never hit the full 100 Mb speed of my network when connecting to a server on the net.

Re:What value?

[michrech]

You also have to account for your neighbor who is addicted to porn and downloads it constantly seeding at 100% for days on end.

Hey, don't talk about me like that when I'm not around ;)

Only two minutes from OP to reply -- You type pretty fast for only having one hand available!

Re:What value?

[geekoid]

But if you look, that whole sentence could be typed with just the left hand keys.

no, not really I just wanted t make you look~

Re:no good

[von_rick]

You can have a repeater every 3 inches. Simple.

Too early?

[NotPenny'sBoat]

Tbps ethernet seems a bit early. Correct me if I'm wrong but isn't the average read on a SATA somewhere around 5 Gbps?

Re:Too early?

[norkakn]

10 Gbps is already normal in server rooms. OC-768 is in the wild at around 40 Gbs. 100 Gbs is definitely around in labs, but I'm not sure if any of it is retail yet.

SATA doesn't have to be very fast, because a single hard drive isn't very fast.

Re:Too early?

[fuzzyfuzzyfungus]

If you have, say, a bundle of fiber running across the Pacific that would cost you 9,334 bazillion dollars and a battle with the giant enemy crab just to upgrade; being able to increase its capacity just by upgrading the hardware on each end is a very attractive proposition. This applies, to a lesser degree, in all but short run situations.

This isn't exactly destined for workstations in the near future(heck, neither is 10GigE, and that is more or less commodity-off-the-shelf stuff by now); but there are applications where higher speed per fiber could well be desirable.

Re:Too early?

[LoRdTAW]

I suppose you forgot about internet back bone links. Terabit Ethernet should hopefully enable Tier 1 ISP's to provide really fat pipes to ISP's so we can finally get more bandwidth. The bigger the backbones the faster our broadband can be. Well at least that's my fantasy. 100mbit boradband should be cake walk with tubes that fat.

Re:Too early?

[norkakn]

Let 1000 people drink?

Re:Too early?

[QuantumRiff]

The article talks about using DWDM to basically multiplex multiple 40Gbps wavelengths on the same fiber. Separating out the wavelengths at the other end is the part where the speed limitation seems to be. 40Gbps has been around for awhile, and so has DWDM.

Re:Too early?

[Anonymous Coward]

No, with modern repeaters, that is not necessary. The repeaters are pure optical amplifiers that don't care how the signal is modulated. Only the end equipment needs to be sophisticated to do the fine-grained wave-division multiplexing - so you don't need to pull up the repeaters to upgrade the capacity. It's really quite neat. See http://en.wikipedia.org/wiki/Optical_amplifier

All-natural ingredients...

[geekmux]

"...a form of the exotic material type, chalcogenide, or arsenic trisulfide.

Whew, for a minute there I was worried we were going to use some hazardous materials.

Re:All-natural ingredients...

[von_rick]

But its got what networks crave, its got electrolytes.

Re:All-natural ingredients...

[drinkypoo]

You laugh (me too) but I'd sure like to know what we're going to do with all the Arsenic we have lying around. I mention it often, but here in Lake County California we have a superfund site full of the stuff. If we could bind it up and then dope something with it that would be very stable, it might give us a future use for the stuff that would let us not dump it into a concrete pit, fill it up, then pave over the pit, build some new walls, and add more arsenic.*

* I don't know that they're actually doing this with Arsenic at The Geysers geothermal power plant, but they are definitely doing it with other toxics. The superfund site is out of Middletown, where Calpine's office is/was located.

Content Filtering

[El Torico]

Too bad my bullshit detector only operates at about 500 words per minute.

Re:Star Trek solutions?

[von_rick]

...I enjoy any solution that uses "photonic" anything and "arsenic trisulfide" anything. Cool

Interestingly, it conjured an image in my mind that is a mix of baby-formula and pesticides.

Still needs work

[smooth wombat]

The solution involves a photonic chip that uses laser light for switching signals, and a form of the exotic material type, chalcogenide, or arsenic trisulfide."

Once you have the photonic chip installed, you will need to realign the deflector shield to output a graviton pulse through the arsenic trisulfide to create an anti-tachyon pulse which will modulate itself based upon the resonant frequency of the transport medium, thus allowing for longer distance transmittal of data than is currently possible.

Granted, it will take 15 years and research team of a hundred to complete, but it is doable.

Re:Still needs work

[ionix5891]

Seven is that you?

Re:Still needs work

[Ragzouken]

Keep dreaming, it's La Forge.

Re:Still needs work

[eggboard]

Okay, I'm the author of the Ars Technica piece, and that make me laugh.

Talking to the researcher, Eggleton, made my head slightly explode, because he's looking 5 to 20 years into the future with the research he's on top of today.

But they have practical devices that show that the stuff can be hand-built, and that's what blows my mind.

The future isn't in plastics -- it's in glass!

It would also appear...

[Brain-Fu]

...that we are inching our way towards the metric system.

Re:no good

[Anonymous Coward]

You can have a repeater every 3 inches. Simple

Re:no good

[MadnessASAP]

You can have a repeater every 3 inches. Simple

Re:no good

[FishWithAHammer]

Obviously you forgot your "NO CARRIER".

Re:Do The Math,..

[Ragzouken]

It's a pity that technology like this never gets cheaper.

Re:no good

[PTBarnum]

The article doesn't say how far they can send the terabit signal, only that the receiver requires 5 cm of fiber to split the signal into lower bandwidth pieces. Presumably the distance between sender and receiver is longer than that.

Re:Star Trek solutions?

[Mister Whirly]

Actually, that is a new Chinese product where the pesticide IS the baby formula. Now with more melamine flavor!

Re:no good

[ebuck]

For such a high speed link, I think that a CSMA/CD technology is probably the wrong answer. Your "bubbles" in the network wire of collision screaming must be incredibly wasteful. But hey, we live in a field where waste is justified by the comparative cost of hardware upgrades over man hours.

I'd love to route Ethernet packets tunneled through an ATM link set up on this kind of bandwith, but somehow I don't think that solution requires cutting edge research.

Ethernet is good at doing what it does well, but the tuning require to make this truly effective might make the mess of jumbo packets look like child's play. I imagine that as the speeds increase, so will the issues. It's a remarkable achievement to push Ethernet this fast, but it seems that it's the equivalent of making your car travel 0.6c. At those speeds you have to wonder if you're using the right vehicle.

Re:no good

[petermgreen]

For such a high speed link, I think that a CSMA/CD technology is probably the wrong answer.
The modern way to build an ethernet network (at least the important parts of one) is to use switches and full-duplex point to point links. Full duplex links do not use CSMA/CD.

CSMA/CD is rarely used at gigabit (I don't think i've ever seen a gigabit hub) and isn't supported at all at 10 gigabit and above.

TFA and the /. summary poorly titled though, this is about a physical layer advancement. That advancement may eventually lead to terabit ethernet, it may also lead to other standards with similar perfomance.

Ethernet is good at doing what it does well
What ethernet has done well is maintain compatibility accross gnerations. If I have old equipment with a 10baseT controller I can plug it straight into a modern network. If I have even older equipment with a BNC or AUI connector I can still connect it to a modern network without too much fuss or expense.