Please create an account to participate in the Slashdot moderation system

typodupeerror

## 2.56 Tb/s Transmission Record213

RalfM writes "2.56 terabits of data per second in new transmission record by Bell Labs, Lucent's research arm." So this thing could transmit my entire mp3 collection in under a half second.
This discussion has been archived. No new comments can be posted.

## 2.56 Tb/s Transmission Record

• #### And what then? (Score:5, Funny)

<saibot&hesperia-mud,org> on Saturday March 23, 2002 @11:40AM (#3212743)
So this thing could transmit my entire mp3 collection in under a half second.

Good, and then you'll have to wait 4 hours for your HDD to write them ;-).

• #### Re:And what then? (Score:1)

Let's say you have 2.5 gigabytes of mp3s.

Most harddrives can write 30 megabytes per second when tweaked with hdparm:

2500 (megabytes)/30 (megabytes/s) = 83.33 (seconds)
• #### Re:And what then? (Score:1)

your math is for only 2.5GB of mp3, but by taco's math he must have over 100GB of mp3 which means way more than 83s
• #### Re:And what then? (Score:1)

Note he said less than half a second - that just means he must have between a 0 and 1.28Tb mp3 collection. All these figures people are coming up for it are just guesses.
• #### Re:And what then? (Score:2)

2500 (megabytes)/30 (megabytes/s) = 83.33 (seconds)
Sure, 83 seconds, for 2.5 GB, but we're talking here about transfer capabilities of 320GB/s, so approximately 11000 times faster than your HDD can write data... I fail to understand why you replied with these figures.
• #### Huh? (Score:2)

Most people I know have way more then 2.5gigs of MP3s, I think I have 20 or so myself. Taco indicated that he had about 150gigs himself. (1.28Tb = 163TB)
• #### Re:And what then? (Score:2)

> Good, and then you'll have to wait 4 hours for your HDD to write them ;-).

This transmits 320GB/s; food for thought is that most system MEMORY architectures would struggle to reach 1/400th of that. Yummy.
• #### Bandwith problems? (Score:1)

If Bell Lab's new technologies are used, does this mean the bandwith problems discussed recently can be fixed?
• #### Re:Bandwith problems? (Score:4, Informative)

on Saturday March 23, 2002 @12:23PM (#3212920) Journal
No, 'cause the LambdaXtreme unit used is unbelievably expensive, and you need at least 2 of them.

Also, you need EBDA single-mode fiber, which isn't the majority in the ground.

Soon, though.

Charles E. Hill
Core Network Engineer
Lucent Worldwide Services
• #### How does this stack up? (Score:1)

I was wondering... How does this stack up next to the transferrates of those fiberoptic telecomunication cables, like the ones they lay underwater and I beleave are used in the net's backbone?
• #### How practical exactly..... (Score:3, Insightful)

on Saturday March 23, 2002 @11:51AM (#3212788)
While I have to admit 2.56 Tb/s is darn impressive the article doesnt mention if this can be applied to currently existing fiber optic networks. After having invested millions in new fiber and equipment for my area I seriously doubt my provider (twc) would be willing to just jump up and make large (read as: expensive) modifications. Especially if our network is "just good enough". Advances like this are interesting but how long will it take to "filter down" to us consumers?
• #### Re:How practical exactly..... (Score:3, Insightful)

Most of these speed records are about as applicable to everyday communications problems as the Thrust SSC's land speed record is to the problems of everyday ground transportaton.

Eventually the technological advances will influence everyday communications technology, but for now they're a gee-whiz thing that's of little direct value.
• #### Re:How practical exactly..... (Score:2)

Actually, Time Warner was one of the first Lambda Router customers last year. The LambdaXtreme unit is what the article mentions and it is a super-long-haul unit. AOL/TW was supposed to be using the 1.6 Tbps version for their coast-2-coast backbone.

• #### Very practical. (Score:1)

Similar stuff has already been deployed and is in use today. Here are two examples. The first is a MCI Worldcom deployment [nortelnetworks.com] from 2001. Your voice calls and data are already flowing around the world on these pipes. The second is, perhaps surprisingly, a Chinese deployment [nortelnetworks.com]. I'm sure that there are others too but, frankly I can't be bothered to look them up. I know, from personal involvement that there are several other high-speed installations around but, these do not span such great distances. Rather the are 30 and 70 mile rings around metropolitan areas but, they are just as fast and in one case even faster.
• #### Re:Very practical. (Score:2)

that wasn't really the question.. I think most ofus know that our data/voice is going through fiber, the question washow expensive is it to upgradecurrent systems to the newest technology.
• #### Re:How practical exactly..... (Score:2)

If your provider [twc] is Time Warner Cable they might.

In my area they went to work faster than a lot of cities. Now they dominate the market and we are the second city to get HBO on demand and we also have another video on demand library.

They own all that content that I er you want to steal so they might actually be interested.

Nice try... but if you wanted to shoot down the practical side you should have pointed out that your computer isn't going to like all that data so fast.

Streaming mp3s, every song... ever!
• #### Ping (Score:3, Funny)

on Saturday March 23, 2002 @11:54AM (#3212806)
Maybee I can finaly get a good ping in quake now.
• #### Unfortunately it would still take 10 minutes.. (Score:1)

to transmit my Natalie Portman jpegs..
• #### Re:Unfortunately it would still take 10 minutes.. (Score:2, Funny)

by Anonymous Coward
This may come as a shock to you, but resizing a 400x300 picture to 400000000x300000000 does not result in a more detailed picture.
• #### Re:Unfortunately it would still take 10 minutes.. (Score:1)

That was awsome... I almost wet my pants when I read that... I just wrote a script to change the size/res on all my Digital Photos to smaller sizes, and you are right... You can get away with much lower res then you would think. For some reason everyone wants the pictures I take, and I was starting to think I was going to need one of those terrabit connections to send them all... How many pitures at 400x300 could you send in a minuet at that speed?... note my lame attempt to bring this post back on topic...
• #### To where? (Score:4, Funny)

on Saturday March 23, 2002 @11:58AM (#3212820)
So this thing could transmit my entire mp3 collection in under a half second.

Sure, and unless you have a storage device that can accept data at that speed, the only place your MP3s are going is /dev/null, so you may as well save the net bandwidth and use the mv command.

• #### Re:To where? (Score:2)

Actually, and I'm serious, I'd be impressed if he could move data to /dev/null that quickly. I don't know of any small computer bus architectures available right now that could handle that kind of speed.

More impressively, almost 4 million people could streem unique mp3's over this line simultaniously.
• #### Re:To where? (Score:2)

This kind of speed would not be for connecting computers to each other directly, but rather big comm backbones. On each end, fast switches would split this out into more manageble rates.
• #### Re:To where? (Score:2)

That's exactly what I'm trying to say.
• #### Re:To where? (Score:1)

Oh yeah... it is... Wish I could retract my comment.
• #### Re:To where? (Score:1)

You are skipping the first problem. Its not to where.?. its from where.?. I'd be impressed if he had a disk/memory, device that could transmit that fast to the network equipment.
• #### well now I am torn (Score:3, Funny)

on Saturday March 23, 2002 @12:01PM (#3212832) Homepage
Is it a Beowulf cluster of these that I want to imagine, or do I want to go for the usual "that's a whole lot of pr0n!" comment?

being trite and obvious has never been harder...

• #### Re:well now I am torn (Score:1)

You could always have a Beowolf cluster of them transmitting pr0n..
• #### Re:well now I am torn (Score:1)

Well, the pr0n comment would allow you to quote the Simsons...
Generic Nerd: "I wrote a program which downloads porn a million times faster!"
Marge:"Why would anyone need that much porn?"
Homer: "soo much..." (drool)

That always gets you karma.
• #### Re:well now I am torn (Score:2)

12:00:00.0 installed 2.56 terabit internet connection.
12:00:00.3 downloaded every existing JPG of Natalie Portman.
12:59:59.9 Put grits on the stove.

P.S.
Ok, she was in a starwars movie, but it was a bad starwars movie. Ok, she's attractive, but she's not amazing. So, why the hell is she this huge running /. routine? And what's with the Hot Grits?

-
• #### Re:well now I am torn (Score:1)

I think it's because she's young, attractive, and extremely intelligent. Kinda geekish too, I believe.

• #### Re:well now I am torn (Score:2)

extremely intelligent. Kinda geekish too

Ahh. I don't know much about her out-of-character.

But, uhhh... the grits? hehe

-
• #### Novels? (Score:1)

...(Sending a gigabit of information per second is equivalent to transmitting the information content of approximately 1,000 novels every second; sending 40 gigabits per second over 64 channels is equivalent to transmitting the information content of 2,560,000 novels)...

Information measured in units of novels? Novel is text only data. So, 1000 novels doesn't sound impressive. 1 DVD per second would be more impressive :).
• #### Re:Novels? (Score:2)

I think this is meant for the "common person" to understand. Of course the "common person" has a really good idea of how a novel, or even a giganovel compares to digital content such as web-pages. So I really think the point of the comparison was to have a Big Number. :)
• #### Re:Novels? (Score:1)

Let's see.. networking and HD units are usually base 10. 1,000,000,000 / 8 / 1000 = 125,000 bytes. This is actually low; most of my ebooks are in the 300-500k range. Maybe they just read juvies?
• #### Re:Novels? (Score:3, Interesting)

No they're not. Grrrr...

Maxtor started the base 10 crap in the mid 90's so stupid users could figure out how much space they had.

64kbps, 128 kbps, 384, 768, 1.54 Mbps...

And file sizes are STILL in traditional base 2. When someone says a file is 1KB, it's 1024 bytes, not 1000 bytes.

64KB of RAM is 65536 bytes, not 64000.

Just because someone bastardized the numbers for idiots doesn't mean it's actually propagated into reality.

Packaging and marketing doesn't change the guts, it just makes it easier for the average joe to feel like s/he knows what s/he's talking about, even if they don't.
• #### Re:Novels? (Score:2)

No, the base two numbers are stupid. They are only marginally reasonable for RAM size, but it makes no sense for harddrive sizes. Can you tell me how many bytes are in 43MB, quickly? Mixing base 2 and base 10 makes no sense because humans always have to convert it to base 10 in the end.
• #### Re:Novels? (Score:2)

Can you tell me how many bytes are in 43MB

What the hell diffrence does it make? I mean really? Are you ever going to need to know the exact number of bytes of that file? (and if you are, are you ever going to need to figure it out by hand?)

The maximum size of anything on a computer is going to be a base2 number, always. two gigs max for a file means two gigs, (2^30). There is no more reason to mesure ram in that size then hd space.
• #### Re:Novels? (Score:3, Interesting)

No, they are. Gtrrr indeed!

Some sources define a megabit to mean 1,048,576 (that is, 220 ) bits. Although the bit is a unit of the binary number system, bits in data communications are discrete signal pulses and have historically been counted using the decimal number system. For example, 28.8 kilobits per second (Kbps ) is 28,800 bits per second. Because of computer architecture and memory address boundaries, bytes are always some multiple or exponent of two. See kilobyte, etc.

So what's that mean? An ISDN 64Kbps B channel is, in fact, 64000 bits per second. A typical 115.2Kbps maximum rate on a PC's serial port is 115200 bits per second.

Hard drives [ata-atapi.com] are also measured using SI definitions. The power-of-2 definitions come from memory. Memory devices often inherently have power-of-two sizes, since the n address bits going into a memory provide 2^n addressable cells. Hard drives have no inherent reason to be a power-of-2 size so SI units make much more sense.

• #### Kosherliciousness (Score:1)

Well, all them terabits are all well and good. But what about ping times? Would this system make them infantesimally small? Or would it ping as bad as satmodems do? 1 mu57 kn0w! 1 n33d 17 70 pl4y my c0un7ar57r13k!

Bah. All I really care about is that 2 terabits/second would download me a whole lotta porn, a whole lotta fast.

Er, no, scratch that. Most servers are on ADSL/T1/T3s, which can only output at a certain, arbitrary, preset rate. So. I ask you, what's really the point? It'd be like having a cablemodem back in the days that everybody's BBSes were running off of 300 baud modems. I honestly don't think that you'd see a terribly higher transfer rate than you already do off of your cablemodem or dorm T1/T3/whatever.

But, baby... Imagine these 2tb lines becoming the standard... drool baby drool.

I still long for the day that I can pull broadband out of thin air. Now THAT'D be sweet...

Heavy heavy fuel... Heavy heavy fuel... if you wanna run cool... you've got to run...
• #### Re:Kosherliciousness (Score:1)

Well.. I guess considering ping times on a 100mbit lan are typically 0.4ms or some such figure, the ping time on such a high speed connection would be 0.1ms or better I would imagine.. no latency whatsoever.
• #### Re:Kosherliciousness (Score:1)

The ammount of data you can cram down the pipe has no bearing on your ping unless you saturate the pipe. And, frankly, Quake packets are TINY. Its the speed of the routers your data has to go through that effects your ping. Why do 56ks suck for gaming? Because they have to convert the signal from digital to analog, and your ISP has to do the opposite. Your nifty cable modem is 100% digital and completely avoids that nasty ms-adding conversion. But you have to go through hardware other than the wires, which is what adds the pings. Put 50 routers on a 100mbit LAN in between 30 computers and the Quake server, and have them all connect and play. See how your pings change.

Whether you have a 1" pipe or a 30ft. pipe, it doesn't matter - a drop is still a drop.
• #### Re:Kosherliciousness (Score:2)

um, I don't think this thing is meant as a "consumer" kind of service.
• #### Re:Kosherliciousness (Score:1)

Give it a few years ... lots of this trickle down to the consumer market eventually.
• #### Re:Kosherliciousness (Score:2, Informative)

Ping times (for small ping sizes) are approximately equal to the amount of time it takes for light to pass through the fiber and back. Most people in telecom use 1ms per 100 route miles of fiber as a rule of thumb. (remember the fiber does not follow a straight line between cities!) So if it is 4200 route miles between NY and LA, your ping will take 42 ms to get there and 42 ms to get back = 84 ms.

You can't change this basic fact of physics. If you were running somehow on a microwave system instead of fiber, you would actually approach the speed of light in free space, so your ping time would be a little faster.

By the way, there are other factors that go into the delay of your ping besides propagation delay. Other things such as: congested buffers in routers along the way, serialization delay to clock the packet out of your device (negligible on fast links, but a big effect on dial-up), the response time of the remote device, devices delaying packets to do an ethernet arp, etc... Propagation delay usually ends to be the biggest factor when you are talking about ping times.
• #### This is not a step forward.... (Score:1)

.... Using the existing fiber that is in the ground would be.

Tom
• #### Re:This is not a step forward.... (Score:1)

If a paper gets accepted for presentation at a conference like this one did, then it is always going to be a step forward.
• #### Re:This is not a step forward.... (Score:1)

Yeah, assume the technology works and its all great and that. Its still a completely different challenge to get a telco to invest in it and actually implement it.

Tom
• #### Re:This is not a step forward.... (Score:1)

These systems work over the existing fiber.

These systems reduces the optical to electrical to optical conversion points along these fibers. This
is what saves the carriers a lot of money.

Additionally, these are both time, and wave division
systems. This basically means it's like transmitting am and fm stations over the wire. (Turns 1 wire into
160).

They also allow routing of the aforementioned stations (lambdas) so that you can dynamically set
up circuits with a click of a mouse.

• #### Can they sell it? (Score:1)

Just this week, I saw that the big news pages are going to start charging for video clips, and the like. The reasoning is that bandwidth costs too much for this kind of service on the web.

What I find interesting is that the optical sector has a ton of equipment which could/should reduce the cost of bandwidth by a tremendous amount, but nobody seems to be buying it.

• #### Re:Can they sell it? (Score:1)

Given that this is Lucent, I think it's pretty much guaranteed that they can't sell it.
• #### So... (Score:1)

On Thursday, we have a slashdot story [slashdot.org] that we're going to face a bandwith shortage RSN and today we have Lucent to the rescue [yahoo.com] Put your buy orders in now!
• #### Impressive (Score:1)

I'm curious about what they used to generate the data. You need some serious fire power to keep a pipe that wide full all the time.
• #### Re:Impressive (Score:1)

The data was probably just simulated using something that can modulate the signal as fast or faster than that pipe could hold. It was probably just a nonsense stream of bits randomly generated, but I'm just guessing.
• #### Re:Impressive (Score:2)

For the testing to be meaningful, you need to be able to check and compare it with what you get at the other end - So they need SOMETHING recording the data, at both ends.
• #### Thats nice. (Score:2, Interesting)

The theoretical maximum (for silica) I've heard quoted is 40Tbit/s, but I'm sure you could squeeze a bit more out. The current limit is the gain spectrum of the Erbium Doped Fibre amplifiers that make sure a signal can travel long distances, these have a (relatively) narrow gain band. Semiconductor Optical Amplifiers (think of them as diode lasers, without the mirrors) could have a wider spectrum than the optical fibre! Lots of problems with them currently though.

I think I'll wait for the quantum dot lasers [google.com] to catch up.
• #### Re:Thats nice. (Score:2)

The theoretical maximum (for silica) I've heard quoted is 40Tbit/s, but I'm sure you could squeeze a bit more out.

Hmmm. I'll tell you how to squeeze more out... take whatever was giving you 40Tbit/s before. Lay down 7 more of those. In parallel. Presto! Now you've got 40TBYTES/s. Repeat as required.

"Buy 8 and I'll throw in a parity bit for free!"
• #### Re:Thats nice. (Score:2)

That's cheating dammit!
• #### Re:Thats nice. (Score:2, Informative)

Actually, the article said they used raman amplification, not Erbium Doped Fiber Amplifiers. Raman can be used in either the C-band (1530 to 1565nm - typically where EDFAs are used), or the L-band (1565 - 1625nm). Actually, I think Raman can be used at any point in the spectrum, but don't quote me on that because I am not a physicist!

EDFAs do allow signals to travel long distances, but the problem is that they are "single point" amplifiers. The signal gets attenuated down quite a bit after traveling 50 miles or so and then is boosted back up by the EDFA. The signal to noise ratio has already dropped too low and can't be recovered. So with EDFAs you are limited to a relatively small number of hops (six or so) before it has to be electrically regenerated (detect the signal and then electrically send it to a laser to be retransmitted cleanly).

Raman amplifiers use an effect called Stimulated Raman Scattering that uses the fiber in the ground as the amplifier itself. By using a pump laser transmitted into the fiber (typically opposite to the direction of data flow), the power of the pump laser's low wavelength is transferred in the glass to the higher wavelengths of the data signals. Amplification is then distributed along that 50 miles of fiber or so between pump lasers. The signal power never gets as low with Raman as with EDFAs, so the signal to noise ratio is kept higher. That's how they are able to get much longer distances between electrical regeneration out of these new systems.

I don't know much about SOAs, but I've heard they are not good for multichannel systems. Something about the fact that they are noisy? I think they are good for single channel applications.

This research is great and all, but telecom providers are having enough trouble selling capacity on their existing EDFA systems. Something is going to have to drive a lot of demand for bandwidth before systems like this one from Lucent get deployed.
• #### Re: That's no record! (Score:1)

I'm sorry, but the record referred to (see the article) is on ...high-bandwidth, ultra long-distance transmission[s].
We're not talking last-mile here, we're talking a distance of 4000 kilometers (2500 miles), roughly the distance between Orlando, Fla., and San Diego (article).
That's a little different, wouldn't you agree?

--
m iso socially aware artistic geek pen-pal, m or f, in '1337 edu. jazz, poetry a must.
• #### Engineer wanted (Score:4, Funny)

on Saturday March 23, 2002 @01:01PM (#3213059) Homepage
So this thing could transmit my entire mp3 collection in under a half second.

Engineer wanted for creation of 2.56Tb/s DRM system. Must be able to scan for copyright flags in data stream and deny transfer permission.
• #### mmmm.... spam (Score:1)

just think of how much more spam could be sent if you had that much bandwidth. Mmmm... streaming dvd quality video spam.

or, replace spam above with...

pr0n...warez...romz...iso ripz...pr0n

i cant wait
• #### I'll take 2. (Score:1)

Yeah... I'll take 2... just put it on my credit card. Having a 2 tb/s would be fun to have even though it transfers information faster than my computer can process.
• #### entire mp3 collection? (Score:1)

this thing could transmit my entire mp3 collection in under a half second.

ummm ya... if you happen to have your mp3 collection in on (many) pc(s) with a memory transfer rate of 2.56 terabits/s and also happen to have a modem that can transfer 2.56 terabits/s!!

• #### that's nothing (Score:1)

That speed is soo slow. Why is this a new record? I mean my TI-89 can transfer at least twice that. geez..
• #### price? (Score:1)

Damn, it must be expensive to setup a line like that. Just the boxes in each end that must deliver the data. It must take some processing power to break the flow down to slower connections, like "slow" gigabit connections.

Anyway what are the requirements for the fibers and how much could you speed up existing lines? I guess it depends on the quality of the fiber.
• #### 2.56Tbps=320Gigabytes/second (Score:2, Interesting)

What sort of device can read that quickly? That's an order of magnitude or so faster than the fastest RAM I've seen. I suspect they simply transmitted a simple repeating pattern rather than actually reading and writing data from a device of some sort.
• #### Re:2.56Tbps=320Gigabytes/second (Score:2)

Probably. Still, its an advance, and with the internet being central to todays computing experience, with this tech on the horizon, its an incentive to keep Moores Law going(in addition to the fact that if AMD OR Intel slacked off the other would eat them alive)

Also, as I said in a previous comment, this would have potentially huge advantages over current network technologies in clustering.
• #### Transmission Record (Score:4, Funny)

on Saturday March 23, 2002 @02:17PM (#3213310) Homepage
And in related news:
These same engineers hope to set a new 1.00 Tb/s reception record later today.

-

*drool*
• #### *blinks* (Score:1)

Even just connecting to some poor web server, itll get slashdotted :D

Not much good on slashdot tho, im still throttled by the 20 second limit on posting on slashdot :D
• #### Taco's MP3 Collec (Score:2)

So this thing could transmit my entire mp3 collection in under a half second.

Most of those gains are due to the following ingeniuos compression scheme:

1. Download Taco's copy of Bobby Vinton's "Melody Of Love".

2. Instruct the client to make 135,275 copies locally.
• #### And this helps me.... how? (Score:2)

...have doubled the distance record for high-bandwidth, ultra long-distance transmission by sending 2.56 terabits (trillion bits) of information per second over a distance of 4000 kilometers (2500 miles), roughly the distance between Orlando, Fla., and San Diego.

Yet, I still can't even get 144 Kilobits/s from Verizon at 5 miles... where's my Fiber-to-the-home?

• #### With Pipes Like this.... (Score:2)

I can see that the media industry (maggots) will have a plan for you to have a 500gb/sec connection and your machine will be a thin client to a server at the RIAA HQ (If it is not already destroyed), Cant to much about copyright 'protection' now could we?

As In, If you want a 500gb/sec pipe you have to run one of our thin clients. This pipe will only work with it. No PCs will be able to use it because of Encryption, authentication etc.....

#### Related LinksTop of the: day, week, month.

The trouble with money is it costs too much!

Working...