Slashdot Mirror
10 Terabit Ethernet By 2010
Eric Frost writes "From Directions Magazine: 'Because it is now impossible to sell networking unless it is called Ethernet (regardless of the actual protocols used), it is likely that 1 Terabit Ethernet and even 10 Terabit Ethernet (using 100 wavelengths used by 100 gigabit per second transmitter / receiver pairs) may soon be announced. Only a protocol name change is needed. And the name change is merely the acknowledgment that Ethernet protocols can tunnel through other protocols (and vice versa).'"
From the article: "iSCSI (Internet SCSI) over Ethernet is replacing: *SCSI (Small Computer Systems Interface..."
iSCSI is far superior to stadard SCSI for obvious reasons, and its widespread adoption will really spark a massive gain in the SAN (Storage Area Network) market. The technology is there, now we just need more major vendors of SCSI devices (especially storage and image filing systems) to make more SCSI devices that support iSCSI natively and applications that take advantage of it. Combined with practical solutions from vendors of network storage software like Veritas we could see some major spending in IT. And more money being spent on IT is always a good thing.
I don't keep up much with the progress of the Ethernet technologies at hand, so is it realistic to suppose that the practical implementation/creation of 100 Gigabit Ethernet, 1 Terabit Ethernet, and 10 Terabit Ethernet will be seperated by merely two years each?
"Because it is now impossible to sell networking unless it is called Ethernet". Incorrect. You can easily sell network gear that is tagged with the "WiFi" designation.
Is there going to be storage that can read/write that fast by 2010 too?
paintball
Bandwidth is good, but what about latency? Ethernet has traditionally suffered from high latencies and doesn't work very well for High-Performance-Computing-Clusters. Myrinet and other ridiculously overpriced networking hardware works much better for clustering. I wish terabit ethernet does something about ethernet latency so that efficient clustering becomes a little cheaper.
What is that article actually supposed to be about? Seems like a scrambled mess of acronymic buzzwords with no actual content to me.
...is there going to be a bus on desktop machines that can read or write that fast?
Probably not. But I could definitely see it being useful for top-end server systems with hugely parallel storage and memory access.
The article is already slashdotted so I can't read it. So what is it refering to? 10Tb LAN speeds? If so - who cares? My existing 100Mb (200Mb switched full duplex) LAN is hardly the weakest link.
10 Terabit Ethernet: from 10 Gigabit Ethernet, to 100 Gigabit Ethernet, to 1 Terabit Ethernet
By: Steve Gilheany
(Aug 27, 2003)
Ethernet Timeline
* 10 Megabit Ethernet 1990*
* 100 Megabit Ethernet 1995
* 1 Gigabit Ethernet 1998
* 10 Gigabit Ethernet 2002
* 100 Gigabit Ethernet 2006**
* 1 Terabit Ethernet 2008**
* 10 Terabit Ethernet 2010**
* Invented 1976, 10BaseT 1990
** projected
Every kind of networking is coming together: LANs (Local Area Networks), SANs (Storage / System Area Networks), telephony, cable TV, inter-city optical fiber links, etc., but if you don't call it Ethernet you cannot sell it. Your networking must also include a reference to IP (Internet Protocol) to be marketable.
Above 10 Gigabit Ethernet lies 100 Gigabit Ethernet. The fastest commercial bit rate on a fiber transmitter/receiver pair is 80 Gigabits per second. Each Ethernet speed increase must be an order of magnitude (a factor of 10) to be worth the effort to incorporate a change, and 100 Gigabit Ethernet has not been commercially possible with a simple bit multiplexing solution, but NTT has solved this problem and has the first 100 Gigabit per second chip to begin a 10 Gigabit system [http://www.ntt.co.jp/news/news02e/0212/021204.htm l]. Currently, Nortel Networks offers DWDM (Dense Wavelength Division Multiplexing) where 160 of the 40 Gigabit transmitter/receiver pairs are used to transmit 160 wavelengths (infrared colors) on the same fiber yielding a composite, multi-channel, bandwidth of 6.4 terabits per second. Because it is now impossible to sell networking unless it is called Ethernet (regardless of the actual protocols used), it is likely that 1 Terabit Ethernet and even 10 Terabit Ethernet (using 100 wavelengths used by 100 gigabit per second transmitter / receiver pairs) may soon be announced. Only a protocol name change is needed. And the name change is merely the acknowledgment that Ethernet protocols can tunnel through other protocols (such as DWDM) (and vice versa). In fact, Atrica has been advertising such a multiplexed version of 100 Gigabit Ethernet since 2001. [http://www.atrica.com/products/a_8000.html] Now that NTT has announced a reliable 100 Gigabit per second transmitter/receiver pair, the progression may be 1 wavelength for 100 Gigabit Ethernet, 10 wavelength (10 x 100 Gigabits per second) CWDM (Coarse Wavelength Division Multiplexing) for 1 Terabit Ethernet, and 100 wavelength (100 x 100 Gigabits per second) DWDM for 10 Terabit per second Ethernet in the near future.
iSCSI (Internet SCSI) over Ethernet is replacing: *SCSI (Small Computer Systems Interface, in 1979 it was Shugart Associates Systems Interface: *SASI), *FC (Fibre Channel), and even *ATA (IBM PC AT Attachment) aka (also known as) *IDE (Integrated Drive Electronics) *see [http://www.pcguide.com], Ethernet is replacing ATM (Asynchronous Transfer Mode), Sonet (Synchronous Optical NETwork), POTS (Plain Old Telephone Service, which is being replaced with Gigabit Ethernet to the home in Grant County, Washington, USA ) [see references from Cisco Systems 1, 2, 3, or 4] [www.wwp.com], *PCI (Peripheral Component Interconnect local bus), Infiniband, and every other protocol, because, as described above, if you don't call it Ethernet you cannot sell it. Everything, in every type of, communications must now also include a reference to IP (Internet Protocol) for the same reason.
At the same time that the transmitter / receiver pairs are getting faster, and DWMD is adding channels, the capacity of fibers is increasing, as is the transmission distance available without repeaters. Omni-Guide [http://www.omni-guide.com/; then click on enter] is working on fibers that "could substantially reduce or even eliminate the need for amplifiers in optical networks. Secondly it will offer a bandwidth capacity that could potentially be several orders of magnitude greater than conventional single-mode optical fibers." Eliminating
Tethernet?
"Teachers leave us kids alone
Gigabit ethernet, and 10 gigabit ethernet both have it in their specs to accomodate 100 ethernet and 10 ethernet. Therefore 10 Tb ethernet will be called 10000000/1000000/100000/10000/1000/100/10Base T for the OTHER technologies included. The chip will be bigger unless its fancy FPGAing with the FPGA code downloaded from the driver.
So to sell it as Ethernet they have to make it compatible as such. Or to make things cheaper, they will have to settle on a different name to sell cheaper 10Tb cards only. Cheaper 10Tb cards will sell more than compatible ones.
"Give orange me give eat orange me eat orange give me eat orange give me you." -Nim Chimpsky
iSCSI bascially takes native SCSI commands, wraps it up (encapsulates it), and sends it over the wire. In other words, you could use a SCSI scanner over a network without having to resort to PC Anywhere or something.
Interestingly enough, if you did it wouldn't be a very big success because the internal PCI or PCI-X bus in the system would bottleneck the interconnects. The NICs would need on-board processors to scale with their enormous bandwidth potential so that they could solve problems like matching checksums and other package management tasks and not have to pound on the system bus so hard.
It wasn't long ago that we really started exploiting video chipsets for rendering graphics, either...
...to prevent the Slashdot Effect?
dinner: it's what's for beer
These high speed DWDM systems talked about in this article aren't designed to be used for LANs or home internet connections - they are meant for high speed backbones that span huge distances (such as across the US or Australia).
They carry mutiple 10Gb/s or 40Gb/s channels on one fibre pair - and these individual channels can be added or removed as necessary, and can be treated independantly. Saying this, 10Gb/s is still a lot, and generally that needs to be broken down into more managable sections, such as gigabit copper ethernet or maybe even 100Mb/s.
It may seem like overkill, but at the core of most networks, there is a distinct lack of bandwidth. Maybe the VOD and video calling predicted 10 years back won't happen on these networks, but more applications are requiring these huge amounts of bandwidth.
An example of this sort of system being rolled out is the Marconi Solstis system in Australia. A very small part of that system was designed by me :)
I am sure packetengines (http://www.scyld.com/network/yellowfin.html) is all over this.
These guys had gigabit routers four years ago when I was helping to set up the AFN (ashlandfiber.com)
Cool to see.. mo'faster is mo'betta
anime+manga together at last.. in real time.
I am an EE major and when I was going to university in the late 80s early 90s everybody was going on how fiber was the future and that we'd run out of capacity on copper RealSoonNow: who'd have thought about 10TERABIT ethernet back then! (heck, I was happy as a clam when my lab upgraded from coax to baseT so the jokers couldn't bring down my box by unscrewing their terminators...)
-- the cake is a lie
Lucent was selling their all-optical DWDM switches (Lambda Series) last year. The LambdaXtreme is a 40 Gbps DWDM unit that uses micro-mirrors (MEMS) for switching. Data is not converted to electricity, but stays as photons the entire route. It is capable of sending data through optical fibers for 1,000 KM *without regeneration* and at 4,000 KM *without regeneration* at reduced (10 Gbps) speeds.
They sold a pair of units (and you have to buy at least 2 or they are useless) to Time-Warner. There is one on the East Coast and one on the West and it forms a major part of their cross-country backbone.
8-10 of the units were sold to Korea (South) for use in wiring up their national rail systems. I also believe NTT DoCoMo (Japan) bought a couple.
This is all last year. Since I'm no longer with that company (layoffs), I no longer get all the product updates. These units were in my product group for install, service and support.
Learning HOW to think is more important than learning WHAT to think.
My apologies for both the recursive quoting and name dropping.
Use Python
I'll stop trolling here after I get this out: stop thinking this has anything to do with your top-of-the-line, supergeekin' Athlon.
:(
This technology is namely meant for backbones, be it on a campus level or as a longer haul backbone. Obviously, your desktop will not need to transfer anywhere near that much data within the next, say, 25 years. If you were using your head while you were reading the (albiet poorly written) article, I wouldn't have to troll.
I just wasted your mod points! HA!
And that assumes that transfer occurs at chip speed, which it doesn't. Assuming a modest clock multiplier of 8 between system bus and chip, that's a 15x overcapacity, even if the entire computer were used to transmit.
-Looking for a job as a materials chemist or multivariat
10Tb/s means
5 million 2Mb/sec compressed video streams
Copy a 250GB drive in 1/4sec
23 thousand streams of 24bit x 1600*1200pix x 75hz uncompressed
1.5k byte packets at 670 million/sec
2 billion x 50 byte packets per sec
port scan all ports on all IPv4 addresses in 20 minutes
Every US resident downloads Metallical's new track in 30 minutes of my http server
And this will all be available at Fry's for a $50 NIC and $30 cable ? When ? I'll hold off buying any new network HW 'till then :^)
Seriously, there are some significant implications here. For 1, you won't need a monitor connected directly to the "fast video card" to get the next fancy 3D graphics features. Memory bandwidth and network bandwidth will be similar meaning that clustered NUMA systems will be interesting. Some of the design decisions we deal with today have been because getting the person close to the computer to improve the experience was a critical factor will disappear.
The article mentions DWDM systems with 100 Gb/s per wavelength. That's bogus.
I am an optical engineer at a 40 Gb/s startup. The jump from 10 Gb/s to 40 Gb/s is huge. Many signal degradations (chromatic dispersion, polarization mode dispersion, nonlinearity,
Compensating for chromatic dispersion, PMD, et. al. requires optical components which DO NOT follow Moore's law. These components are handmade specialty devices.
While a business case can be made for 40 Gb/s, the jump to 100 Gb/s is commercially pointless. If you are building a DWDM system anyway, just spread the same data across more 10 Gb/s channels.
What the hell is "Directions", anyway? It sounds like sci-fi fluff meant to entice VC's.