Where's Our Terabit Ethernet?

carusoj writes "Five years ago, we were talking about using Terabit Ethernet in 2008. Those plans have been pushed back a bit, but Ethernet inventor Bob Metcalfe this week is starting to throw around a new date for Terabit Ethernet: 2015. He's also suggesting that this be done in a non-standard way, at least at first, saying it's an opportunity to "break loose from the stranglehold of standards and move into some fun new technologies.""

Misleading name, "Ethernet".

[pla]

"He's also suggesting that this be done in a non-standard way"

No, he suggested that five years ago

We don't yet have the technology described (wave division multiplexing) in our homes because very, very few of us want to bother with fiber in our homes at all.

You can push an amazing amount of data over glass, no one would claim otherwise. You can't, however, drape it across the floor and up the stairs to your switch for a quick LAN connection... Not only does terminating a fiber suck, the first time the dog steps on that little yellow wire, end of connection. By contrast, I've used Cat5 as a structural material (tied a PC to a hook on the ceiling with it) WHILE USING IT for data.

So no, we won't see terabit ethernet anytime soon, unless someone figures out a way to push it over copper.

Bob Metcalfe, hater of open source

[Bogtha]

Has this guy done anything relevant in the past couple of decades? Here's a choice quote [infoworld.com] of his:

Unix and the Internet turn 30 this summer. Both are senile, according to journalist Peter Salus, who like me is old enough, but not too old, to remember. The Open Sores Movement asks us to ignore three decades of innovation. It's just a notch above Luddism. At least they're not bombing Redmond. Not yet anyway.

The hard part of being down on Linux and the Open Sores Movement is worrying about that menace hanging over us at year's end. No, not Y2K, but Linux's nemesis, W2K, Windows 2000, the operating system formerly known as Windows NT 5.0.

W2K is software also from the distant past -- VAX/VMS for Windows. But it will overpower Linux. NT, now approaching 23x6 availability, is already overpowering Linux. NT and NetWare constitute 60 percent of server software shipments. All Unixes make up 17 percent, and Linux is a small fraction of that. When W2K gets here, goodbye Linux.

Wait... you believe Metcalfe WHY?

[1336]

As in the Robert Metcalfe whose Wikipedia article [wikipedia.org] has an "Incorrect predictions" section listing where he wrongly thought that "the internet would suffer a catastrophic collapse" in 1996 and this gem:

Metcalfe is also known for his harsh criticism of open source software, and Linux in particular, predicting that the latter would be obliterated after Microsoft released Windows 2000:

The Open Source Movement's ideology is utopian balderdash [... that] reminds me of communism. [...] Linux [is like] organic software grown in utopia by spiritualists [...] When they bring organic fruit to market, you pay extra for small apples with open sores - the Open Sores Movement. When [Windows 2000] gets here, goodbye Linux.

Just because he did something really cool 35 years ago doesn't make him an expert on related matters now.

It's about Shannon's law too.

[Z00L00K]

For everyone that has been working with communication since the early datacom ages Shannon's law [wikipedia.org] has been important. It's still important, and it means that you can't just push everything through, you have to consider the media used.

In a way it can be tweaked a bit, and that has caused a confusion among those that aren't well into the technological difference between Baud (modulation changes per second) and BPS (bits per second).

Anyway - The classical phone modems can have a speed up to 56kbps, but effectively they stay at 28 to 33kbps. And that on a line that actually only provides 3kHz bandwidth. The trick is that in the 3KHz bandwidth you can have a carrier with less than 3000 modulation changes per second, often 2400. In each modulation change you not only have one bit transferred, but multiple bits. This is achieved by having a variation in both phase and amplitude of the signal.

So to utilize the cabling at the extreme speeds that a terabit Ethernet is you may have to resort to the same technique.

There have also been other techniques in use like using multiple carrier frequencies, like what the Telebit Trailblazer modems did. That technology was very resilient to interference compared to the CCITT standards, but it had other disadvantages instead.

Re:Stranglehold?

[KublaiKhan]

Not entirely.

There are still a few token rings and other such mesozoic cruft wandering around in the wild out there, but they still work--because some clever folks invented a way to get from one kind of network to another.

Keep in mind, also, that it's really only the early adopters--those who are willing to buy 1st-generation equipment--who would get 'screwed over', and they have, by definition (as the first generation of a given kind of thing is always several times more expensive than the 'production' generations), the money to waste on this sort of thing.

Re:Misleading name, "Ethernet".

[ChrisA90278]

I disagree. I bought a spool of fiber for some users who needed to deploy a temporary network and then roll it back up and use it again later. We bought a one kilometer roll on a wooden spool and they would just as you say pull it down stairwells through doors and toss it up in trees. Once they hung it over a freeway in Germany from some utility poles (had to hire local linemen for that one) and then after a few day rolled it back.

I told the fiber cable sales guy I was going to test their sample by placing it in the parking lot and letting cars drive over it for a while. The cable was tough basically it was a bundle of kevlar around a thin fiber strand. The kevlar absorbed all of the abuse. After all they lay fiber cable in the ocean. If it can take being dumped off a ship into the ocean it can take a dogs stepping on it. The trick is to specify the correct cable and don't just buy whatever is cheapest.

I will Settle For 1Mbps

[s31523]

Forget terabit ethernet. I will settle for full, actual 1Mbps (10,100, 1000, etc.) speed for both transmit and receive. Even on my home network, I rarely get full %100 utilization on my LAN. Some PC's are linux, some are Windows. Neither machine ever really reaches its full potential. I looked at other networks as well, even my work LAN, and they have similar issues. I am not a network guru and don't want to spend the time tweaking and configuring. The damn Gbps NIC and network drive I bought should just plug and go and I expect to see speeds reasonably close to 1Gbps, but nope. I see like 1% utilization. Seriously, lets make current technology work as advertised before we start claiming super-fast speeds on other vapor-ware technology. Please?

Re:Who needs it?

[Anonymous Coward]

Jakob Nielsen once famously stated that the "perfect user interface" will require 1Tbps [useit.com].

Of course I live in America so if I got 2Mbps through my broadband connection I'd be happy.

Re:Bob Metcalfe, hater of open source

[anticypher]

I had the pleasure to work on projects associated with Metcalfe at the beginning of my career, notably the migration of Ethernet I to Ethernet II standard. He was an autistic, anti-social, self-centered, egotistical curmudgeon from the start, and despite those charming qualities he nevertheless adopted an ivory-tower academic approach in his later life of hating anything created since his 15 minutes of brilliance.

He can always point to DJB as a worse curmudgeon, so there is that solace in knowing he isn't the most disrespected hasbeen still seeking the limelight.

the AC

I don't think a smattering of emoticons in this post will stave off the imminent hater responses, and there isn't really anything I'd put a smiley to.

Re:Stranglehold?

[GooberToo]

Besides, it's not like this is going to affect TCP or IP or whatnot--this is way down at the bottom of the OSI model at level 1.

Early research indicates IP protocols will not scale well with high speed links. CPU load goes through the roof and because of limited buffer sizes relative to line speeds, retries and fallbacks plague applications. The end result is a slow, high speed link.

In a nut shell, for high speed links to become useful to a large category of users, IP, and especially TCP must be revamped. Some research has already progressed down this road but late I heard, much more is required.

Re:For those of you playing at home, a TB is

[Slim Backwater]

RAID-5 is EVIL! Never Ever, EVER use RAID-5, You will LOSE DATA! RAID-1 or RAID-10 only for production use.

If you are thinking about Raid-5, forget it, just stripe your drives in a RAID-0 and enjoy the performance benefits and keep frequent good backups and test your restores.

Why RAID 5 stops working in 2009 [zdnet.com]
Why aren't disk reads more reliable? [storagemojo.com]
End of Raid 5 [c0t0d0s0.org]

finally, BAARF - Battle Against ANY Raid Five BAARF [baarf.com]

HTH, HAND, don't cry.