Design Con 2015
Breaking News
Blog

400 Gbit Ethernet: The Next Leap

NO RATINGS
View Comments: Newest First | Oldest First | Threaded View
Page 1 / 2   >   >>
Bert22306
User Rank
CEO
Re: Why not an easier path to 1Tb/s
Bert22306   8/18/2014 9:43:54 PM
NO RATINGS
The use of 400Gbps is really an interim technology until 1Tbps links are developed. Of course, you could argue that every technology is animterim until something better comes along.

Understood the point about interim step, and that ultimately everything is an interim step. But as you see in my previous post, SONET/SDH has never incremented by speeds of 4X, and that 400 G speed should not be in the sequence of speed increments.

The 40 G step was motivated by a stop up the ladder of SONET/SDH speeds. There is no stop at 400 G, up that same SONET/SDH ladder, if things progress as they have in the past. If you create a new step of the ladder, why not choose one that's more suitable to Ethernet?

Bert22306
User Rank
CEO
Re: Why not an easier path to 1Tb/s
Bert22306   8/18/2014 9:36:05 PM
NO RATINGS
There is still a desire by many carriers to use OTN as the lower layer transport and keeping Ethernet aligned (as a multiple of 4) makes it nice to have an OUT container that an Ethernet payload will fit into.

Okay, so here's the point I was trying to make. If the motivation is to use a SONET/SDH container, for Ethernet carriage on the WAN, which I think we have agreed here is a motivation, then is there a SONET/SDH container expected to be at 400 Gb/s?

To understand the progression of SONET/SDH, I find it easiest to use the SDH levels, and then compute the SONET equivalent. This makes it more obvious why certain STS levels have been called "dormant." SDH levels go up in this sequence: 1, 4, 16, 64, 256, 1024, 4096, ... In other words, you multiply each SDH STM level by 4, to reach the next speed increment. STS refers to SONET convention, STM to SDH, and the two seem to standardize on the same speeds. So here is the progression:

STS-3 = STM-1 = 155.52 Mb/s

STS-12 = STM-4 = 622.08 Mb/s

STS-48 = STM-16 = 2,488.32 Mb/s

STS-192 = STM-64 = 9,953.28 Mb/s.

STS-768 = STM-256 = 39,813.12 Mb/s

STS-3072 = STM-1024 = 159,252.48 Mb/s

STS-12288 = STM-4096 = 637,009.92 Mb/s

Where is that 400 G?

So, if the sequence of SONET/SDH speed steps is not being followed anymore, after the 40 G level, which I can understand, why not instead choose speed steps that make more sense for the traditional Ethernet speed increments?

40G was an Ethernet anomaly, motivated by STS-768/STM-256. What's the excuse for 400G? Multilane Ethernets in the past have been based on either 10X pipes, or on 2.5/25/250X pipes.

FWIW, the next speed close to 40 is STM-262144, at 40.768 Tb/s!

JSMITH1972
User Rank
Blogger
Re: Why not an easier path to 1Tb/s
JSMITH1972   8/18/2014 3:40:25 PM
NO RATINGS
This goes back to commercial and technical feasibility. There are many issues just moving to 50Gb/SERDES, one of the major considerations is that, even today, there is no layer instrumentation to measure 50Gb/s signaling very well. Another example is just the move from 10Gb/s SERDES to 25Gb/s SERDES required the use of a next generation, more expensive printed circuit board material called Megtron 6. This is 2-3x more expensive than FR4. Moving to even higher SERDES might mean the use of Teflon-based materials at (10-15x more cost than FR4) or optical circuits which can be even more expensive than Teflon –based materials

JSMITH1972
User Rank
Blogger
Re: Do we really need that 400G ?
JSMITH1972   8/18/2014 3:37:53 PM
NO RATINGS
True! Facebook, Google and others are leading the demand to higher bandwidth.  400GE can be built today (relatively speaking) using existing key components that are also commercially feasible (i.e. have material costs that result in affordable Ethernet products). Jumping to 1TB any time before about CY2020 is not commercially feasible. It is important to remember that the IEEE develops standards that have to meet all five criteria:  Broad Market Potential, Compatibility, Distinct Identity, Technical Feasibility, and Economic Feasibility. 400GE can meet all 5, 1TB in the present day cannot.

JSMITH1972
User Rank
Blogger
Re: Why not an easier path to 1Tb/s
JSMITH1972   8/18/2014 3:36:05 PM
NO RATINGS
There is still a desire by many carriers to use OTN as the lower layer transport and keeping Ethernet aligned (as a multiple of 4) makes it nice to have an OUT container that an Ethernet payload will fit into.

JSMITH1972
User Rank
Blogger
Re: Why not an easier path to 1Tb/s
JSMITH1972   8/18/2014 3:27:16 PM
NO RATINGS

Yes. 400Gbps can be viewed as interim until 1Tbps. 400Gbps is viable today given current technology and the economy of components.

MeasurementBlues
User Rank
Blogger
Re: Why not an easier path to 1Tb/s
MeasurementBlues   8/13/2014 5:24:42 PM
NO RATINGS
The use of 400Gbps is really an interim technology until 1Tbps links are developed. Of course, you could argue that every technology is animterim until something better comes along.

Bert22306
User Rank
CEO
Re: Why not an easier path to 1Tb/s
Bert22306   8/13/2014 5:08:51 PM
NO RATINGS
The multiples of 4 (40, 400) is a leftover from telecom speeds. OC-768 is 40Gbps.

In truth, the only 4 related to SONET/STM was 40G. (Parenthetically, the scheme that used 4 and 16 was Token Ring, but that's neither here nor there.)

SONET/STM carriage is not a real issue anymore, as it might have been ca. mid 2000s and prior.

http://features.techworld.com/networking/1294/could-wan-ethernet-replace-sonet-and-atm/

http://www.tccomm.com/Literature/Default.aspx/Ethernet-Network-White-Papers/SONET-to-Ethernet-comparison

http://www.itproportal.com/2014/06/11/beyond-the-city-limits-how-ethernet-is-exceeding-the-metro-boundary/

The motivation to use SONET/STM at the physical layer has gone away, so I'm not sure why this 40G vestige from those days is being perpetuated by multiplying it by 10.

You'll note, even in the article, they're talking about 10 Tb/s in the future. Which says to me that the powers of 10 convention for Ethernet is still in people's minds.

MeasurementBlues
User Rank
Blogger
Re: Do we really need that 400G ?
MeasurementBlues   8/13/2014 9:07:57 AM
NO RATINGS
I recall attending the OFC (optical fiber) conference where a network engineer from facebook told an audience how they needed 100Gbps now (5 years ago) and they wanted 1Tbps speeds.

If someone is willing to pay for higher speeds, someone will develop it.

"And here we are scratching our heads as to how to go to next level of the bandwidth, the next level of storage, the next levl of data centers.."

MeasurementBlues
User Rank
Blogger
Re: Why not an easier path to 1Tb/s
MeasurementBlues   8/13/2014 8:57:21 AM
NO RATINGS
The multiples of 4 (40, 400) is a leftover from telecom speeds. OC-768 is 40Gbps. Once data rates hit 40Gbps, they are now increasing in multiples of 10 instead of multiples of 4.

Ethernet has always gone in multiples of 10.
Optical Carrier:
OC-1 51.48Mbps Optical Carrier SONET - All OC-n are 51.48*n Mbps
OC-3 155.52Mbps Optical Carrier SONET
OC-12 622.08Mbps Optical Carrier SONET
OC-48 2.4Gbps Optical Carrier SONET
OC-192 9.6Gbps Optical Carrier SONET
OC-256 13.1Gbps Optical Carrier SONET
OC-768 40Gbps Optical Carrier SONET


Page 1 / 2   >   >>
Most Recent Comments
Flash Poll
Like Us on Facebook

Datasheets.com Parts Search

185 million searchable parts
(please enter a part number or hit search to begin)
EE Times on Twitter
EE Times Twitter Feed
Top Comments of the Week