March 28, 2011
10:58 AM EDT
by
100GbE standards Q&A with Ciena expert Pete Anslow
I've covered the topic of 100G quite a bit on this blog, which should be relatively obvious considering Ciena's large lead on the competition with 100G products. As 100G networks begin what is expected to be a broad global deployment, the topic of 100 Gigabit Ethernet (100GbE) has raised its head as the availability of 100GbE networks and services starts to become reality.
In the last month, Verizon has highlighted its 100GbE link between Paris and Frankfurt, and Internet2 has announced the ongoing deployment of a 8.8Tbps total capacity network with at least one eye on the need for 100GbE links between research institutions (see Chris Robb’s nice post on the Internet2 blog here).
While I consider myself relatively educated on the topic of 100G transport, I can't say the same for 100GbE. Luckily, Ciena is home to one of the industry's most involved experts in 100G and 100GbE standards development in Peter Anslow. Pete has over 30 years experience in the optical space, and is an active member of IEEE 802.3 and ITU-T SG15.
Pete was kind enough to walk me through the topic of 100GbE and give some of his opinions on where the market may be headed. Below is our conversation:
----------------------------------------------------
Q: You are involved in quite a few standards bodies. Which groups do you participate in and how much time is needed to devote to related group activities and standards development?
A: I am currently active in the IEEE 802.3 Working Group (was Chair of the Optical track of P802.3ba and am now Editor of P802.3bg) and ITU-T Study Group 15 (Associate Rapporteur - vice chair - of Q6 which does the optical transport standards). I am also a Liaison officer between IEEE 802.3 and ITU-T SG15 on the topic of 40 and 100G. These two groups will have 11 weeks of formal face-to-face meetings this year, but (particularly with IEEE) there are also many informal consensus-building meetings via conference call.
Q: There are 100G standards in various stages of development across different standards bodies, including 100GbE in the IEEE. Is this simultaneous development a help or a hindrance in your opinion?
A: I think it has been a great help. The process of developing a standard such as 100G takes a long time. It was 4 years from the 100G Call For Interest (CFI) to approved standard in IEEE. During that time ITU-T also developed the standard for OTU4 to carry 100GbE so that as the initial 100GigE modules became available, it was also possible to transport 100GbE in long haul links. If the two bodies had not cooperated so closely there would have been a period of time when standards compliant 100G transport was only available over very short distances.
Q: This last summer, the IEEE ratified the IEEE 802.3ba standard for 40GbE and 100GbE. Is that "mission accomplished" or is there still more work to be done around refining or expanding the standard?
A: There is still work going on to expand the 100GbE standard. While IEEE Std 802.3ba includes 40GbE over a backplane, there is no corresponding backplane version for 100GbE. This is being addressed by a new “100Gb/s Ethernet Electrical Backplane and Twinaxial Copper Cable Assemblies Study Group” (see http://www.ieee802.org//3/100GCU/index.html). There is also some informal discussion going on about standardizing a second generation of lower cost 100GbE modules without the 10 to 4 lane gearbox that current modules contain. One option for doing this would be to define a four by 25G electrical interface. The timeframe being discussed for starting this effort is a CFI in July 2011.
Q: There are many Carriers that have been involved in the development of these standards. In your conversations with them, where do you think their initial drivers are for deploying 100GbE? Are there any existing roadblocks to deployment? And where might the demand move to in the future?
A: I think that there are three inter-related reasons to move to 100G transport. As Core Routers start to be deployed with 100GbE interfaces it becomes very attractive to map each 100GbE signal straight in to a 100G wavelength using OTU4 - using an inverse multiplex strategy is difficult to manage and uses up wavelengths very rapidly. 100G transport solutions are available with the same 50GHz channel spacing as 10G transport, thereby increasing the capacity of each installed fibre by a factor of 10. Finally, the cost per bit of 100G transport will clearly win out over the lower rates, even though 10G costs are continuing to fall.
I don’t think that there are any roadblocks to 100G. There were big concerns over how difficult it might be to engineer optical links to be 100G capable, but the advent of coherent solutions has demonstrated that this need not be an issue at all. As the market for 100GbE modules matures, I think we will see 100G transport become dominant.
I also think that there is no limit to demand in the foreseeable future as long as transport providers can continue to deliver a reduced cost per bit for higher and higher capacity channels.
Q: There is an ongoing industry debate around the demand and need for 40G vs 100G on the transport side. Is there the same debate with 40GbE vs 100GbE?
A: Yes there is, although the details are different for the client side. Currently, the 40GbE modules are based on mature 10G per lane technology while the 100G single mode solutions use 25G per lane. This has resulted in a significant investment in 25G technology being needed for 100GbE which has to be recouped. IEEE chose 25G lanes because they believed that it is a rate that can become cheap in the near term, but this has yet to happen. The result is that the 40GbE modules today are cheaper per bit than 100GbE ones.
In the longer term, this situation should disappear, but there is expected to still be a significant market for 40GbE in situations such as servers where the host is not capable of generating 100G of throughput.
Q: There are several different modulation formats in the industry for 100G transport. Does the underlying "flavor" of 100G matter for 100GbE traffic?
A: Absolutely not. All Ethernet cares about is receiving all of the bits in the right order at the far end of the link. In fact, there are many more line side modulation formats in use at 40G than are currently being talked about for 100G. However, lots of the 40G formats don’t have the performance required to be practical for use at 100G and beyond.
Q: We're just now rolling out 40GbE and 100GbE, but people are already talking about what's next. What's your view?
A: There are two obvious candidates for the next rate of Ethernet – 400GbE and 1TerabitE. 400GbE looks like it could be implemented as an evolution of 100GbE technology – 16 lanes of 25G in an integrated module. For 1TerabitE you would need 40 lanes of 25G which seems too many lanes to be cheap or something like 20 x 50G, 10 x 100G etc. which requires a technology breakthrough. The other consideration is that on the line side 400G per wavelength is looking feasible with an improved spectral efficiency compared to 100G on 50GHz grid, but it is somewhat doubtful that 1T per wavelength can further improve on this while maintaining adequate reach. Consequently, while I expect that there will be plenty of people saying that 400G is too slow and that there is a need for 1TerabitE, I think that 400GbE is a more practical target for the next rate.
Post A Comment
-
In the Lab with the Ciena 6500’s new packet features
May 22, 2013
-
Ciena 6500: Extreme Packet Makeover Edition
May 14, 2013
-
It’s Here: Ciena unveils its Trans-Pacific 100G WaveLogic 3 submarine transponder for submarine networks
April 23, 2013
-
New P-OTS survey puts Ciena ahead of our biggest competitors
March 18, 2013
-
Coherent optical turns 5: Here’s how it was made
March 15, 2013
-
Take better business decisions with GeoAnalytics
February 28, 2013
-
7 Reasons 2013 is the Year of the Network
February 26, 2013
-
Network Monetization: Build the business of your network
February 20, 2013
-
What is a GeoMesh Network?
January 24, 2013
-
How much bandwidth does Broadcast HD video use?
September 26, 2011 -
Ciena 6500: Extreme Packet Makeover Edition
May 14, 2013 -
Delivering on fully open SDN for service providers
April 25, 2013 -
It’s Here: Ciena unveils its Trans-Pacific 100G WaveLogic 3 submarine transponder for submarine networks
April 23, 2013 -
Ciena's little 3902 Carrier Ethernet switch wins big Cable award
May 13, 2013 -
Infographic: You've had a cable cut...so what?
April 24, 2013 -
On the hunt for 100G OpenFlow apps
May 9, 2013 -
Openness, from the top down: SDN for service providers
April 15, 2013 -
Coherent optical turns 5: Here’s how it was made
March 15, 2013 -
Progress with OpenFlow: Much more than SDN-washing
May 6, 2013
