Service Providers

What’s Next for 100 Gigabit Ethernet?

by Greg.Hankins on ‎01-13-2012 05:26 PM (2,053 Views)

2011 was a very exciting year for us, as we started shipping the industry’s first 2-port 100 GbE card for our Brocade MLX Series routers last summer.  We saw a lot of demand for 100 GbE from customers in a variety of different market segments: service providers, data center operators and research institutions –  to name a few.  In fact, we just announced that we shipped our 100th 2-port 100 GbE card well before the end of last year (you can read our press release for more details).  Last year was also a great year for 100 GbE in general.  Since the IEEE 802.3ba standard was approved on June 17, 2010 we’ve had shipping 1st generation media, test equipment, router interfaces, and optical transport gear on the market from a number of different vendors.  The 10x10 MSA also published three standards in 2011 that define lower cost alternative solutions to the IEEE long reach standards at 2 km, 10 km and 40 km.  This means that 100 GbE is definitely ready for prime time, as it’s now become a mature interoperable technology with broad vendor support.


What’s next for 100 GbE?  Just as we saw with 10 GbE, where the technology evolved over several generations as new components became available, we also expect 100 GbE to evolve over multiple generations.  Right now we have some fundamental 1st generation 100 GbE component technology constraints that limit higher density and lower cost modules.  Today’s CFP module is large, has complex components and uses quite a bit of power.  This is because internally 100 GbE uses multiplexed lanes of different electrical and optical signaling types in order to run the interface at 100 Gbps.  The electrical signaling inside the router that goes between ASICs and the CFP module uses 10 x 10 Gbps signaling lanes known as the 100 Gbps Attachment Unit Interface (CAUI).  The optical signaling outside the router, what the CFP module transmits over the physical media, also uses internal multiplexing.  100GBASE-LR4 and 100GBASE-ER4 use 4 x 25 Gbps wavelengths, and the 10x10 MSA optics modules use 10 x 10 Gbps wavelengths, which are multiplexed over a single pair of single mode fiber.  Multiplexing lower speed lanes to get a 100 Gbps signal is the most cost effective and efficient way to deliver the technology today, but as we have higher speed components available we can use higher speed lanes that make the 100 GbE optics cheaper, smaller and use less power.

To address the requirement for higher density and lower cost, several 2nd generation 100 GbE technology projects have started that we expect to see on the market in 2013+.  These 2nd generation projects are all based around a new 4 x 25 Gbps electrical signaling standard that will enable us to change the electrical signaling from 10 x 10 Gbps to 4 x 25 Gbps lanes between the ASICs and 100 GbE media modules.  For now, the most efficient and cost effective optical signaling still uses 4 x 25 Gbps wavelengths.  The next technology evolution will give us 4 x 25 Gbps signaling on both the electrical and optical side, which will greatly simplify the 2nd generation media modules.  The diagram below shows a simple comparison of the 1st and 2nd generation 100 GbE technologies.


Here’s a little more detail about the new 2nd generation projects that have begun.  The IEEE P802.3bj 100 Gb/s Backplane and Copper Cable Task Force was started in September, 2011 and is working on two new short reach objectives:
•    100GBASE-KR4: 4 x 25 Gbps over >1 m backplane

•    100GBASE-CR4: 4 x 25 Gbps over >5 m copper twinax cable


To develop new optical standards, the IEEE Next Generation 100Gb/s Optical Ethernet Study Group was started in July, 2011.  They are working on two new objectives and are also defining the electrical signaling interface to the next generation 100 GbE media modules:

•    100GBASE-SR4: 4 x 25 Gbps over 100 m OM3/OM4 MMF

•    100GBASE-FR4: 4 x 25 Gbps over <2? km SMF

•    CAUI-4: electrical signaling to the CFP2, CFP4 and QSFP28

Two new optics modules, the CFP2 and QSFP28, will be available for 2nd generation 100 GbE which are shown to the left in comparison to 1st generation modules.  Both will offer higher density and lower power consumption, and will be competing for the highest port density that we can fit onto a card as each is at least ½ the width of the CFP module.

Lastly, as a reference I put together this handy table that lists the current and developing 100 GbE standards (click to enlarge it).  It’s a bit dense, but I wanted to keep everything on one page together so that it’s easier to compare.  If you need a quick primer on the CFP modules that are available today, take a look at my blog on 100 GbE CFP modules that tells you everything you need to know.

What can we expect for 100 GbE this year?  As the 1st generation market grows and benefits from economies of scale and lower cost, I’m sure we’ll see a lot more 100 GbE adoption this year as even more vendors develop 100 GbE solutions.  It will be exciting to see how 100 GbE enables network operators to do new things they simply weren’t able to do with their 10 GbE networks.

In my next blog I’ll be writing about the 100 Gbps Ethernet and optical demo that we’re planning at the NANOG 54 conference.  So stay tuned to the Brocade Blogs for more news on 100 GbE and other service provider topics.

For more information on Brocade’s high density 100 GbE solutions, please visit the Brocade MLX Series product page.


CFP image courtesy of Finisar.

Module diagram courtesy of Molex.

How can we improve the Community?