Ethernet services

Is OTN Overkill?

NEW YORK -- Ethernet Expo Americas 2011 -- The Optical Transport Network (OTN) is about to be run out of town, if Sten Nordell, CTO of Transmode Systems AB , has his way.

During a spirited session on packet-optical transport Tuesday afternoon, Nordell went on a verbal rampage against OTN, saying the technology just gets in the way when applied to an all-packet service.

His point was that because OTN is a digital wrapper, packing all sorts of traffic into a homogeneous form, it obscures the visibility you'd want with pure Ethernet services.

"If you want an SLA on an Ethernet packet and you put it on an OTN frame and you want to look at it -- you can't see it," Nordell said. "The only way to do it is stop the OTN, de-wrap it, look at it ... What you've done is broken your OTN service."

Nordell also mentioned the issue of latency, arguing that the application of a digital wrapper can add delays to traffic transit times. Heavy Reading analyst Sterling Perrin later told Light Reading he wasn't so sure that was the case, but in any event, Transmode is claiming to have latency down to 1.9 microseconds one-way, "lower than most OTN switches can reach," Nordell said in his talk.

The other panelists didn't exactly lift Nordell on their shoulders after this. Chip Redden, senior director of marketing with Ciena Corp. (NYSE: CIEN), brought up the most obvious counterargument: "Not every service we have is packet-based today," and OTN, being a time-division multiplexing (TDM) technology, is suited for carrying other TDM traffic.

Nordell questioned the logic of defining next-generation networks based on old TDM requirements, though, bringing up the fact that OTN (a.k.a. International Telecommunication Union, Standardization Sector (ITU-T) G.709) dates back to last century.

Applying OTN means putting "technology defined 15 years ago onto an architecture for the next 20 years that's going to be 99.99999 percent packet," Nordell said, maybe with a touch of hyperbole. "I think that's a stupid idea."

It would be better to put TDM traffic -- assuming it's the vast minority -- onto its own wavelength, separated from the all-packet majority, he said.

Other panelists came to the defense of OTN in other ways. James Anthony, who plans packet-optical strategy for Fujitsu Network Communications Inc. , pointed out that 100Gbit/s wavelengths will need to be filled; OTN can groom traffic to pack those connections thoroughly. And Steve West, CTO of Cyan Inc. , said OTN provides bit-error-rate correction and forward-error correction (although, admittedly, these things can be had without OTN as well).

Despite his rant, Nordell did admit OTN has its place in the network. In fact, it seems pretty well established that carriers want to see OTN in the network core, Perrin told Light Reading. But the metro network is where OTN's presence is up for debate, he said.

— Craig Matsumoto, West Coast Editor, Light Reading

Page 1 / 2   >   >>
menahemk 12/5/2012 | 4:49:25 PM
re: Is OTN Overkill?

On another event this year Sten was violently against TDM, both OTH & SDH, and against Circuit Emulation as well. Along he explained that Transmode products convey circuits and packets. Shall we assume proprietary mapping?

photon2 12/5/2012 | 4:49:24 PM
re: Is OTN Overkill?

A typical rant from Stan again.  This has been his mainstay for years now.  Actually, it's getting boring.


Pete Baldwin 12/5/2012 | 4:49:23 PM
re: Is OTN Overkill?

The panel I wrote about here didn't discuss submarine networks, but that's a place for OTN.

Shankar Narayanaswamy of Reliance Globalcom is speaking at an EENY now and just said he absolutely needs OTN for submarine. An undersea link has a limited number of wavelengths and you're not going to be adding more very quickly. And it has to support every type of service imagineable -- STM64, STM3, GigE, wavelengths... so, he wants OTN, for packing these things into wavelengths.  Pretty obvious use case.

Pete Baldwin 12/5/2012 | 4:49:23 PM
re: Is OTN Overkill?

Sten gets incendiary on purpose; it's part of his style and, yeah, it makes for more memorable panels and good headlines. (Once, anyway.)

Between breaths he did say OTN has its place. Transmode isn't kicking it to the curb quite as much as he was implying.

And while I understand that Sten's style might get tiresome after repeated listenings, it did the job: The audience stayed engaged in a late-afternoon session, and some people were still talking about the panel this morning. I'd be glad to see him appear on a EENY 2012 panel.

neyo 12/5/2012 | 4:49:20 PM
re: Is OTN Overkill? ODUflex together with a control plane protocol such as GMPLS will resolve the issues raised by Sten. The ODUk containers will be dynamically allocated based on the packet traffic demands. IP data offload is a popular use case for OTN.
Jon B 12/5/2012 | 4:49:19 PM
re: Is OTN Overkill?

Hi Neyo,

No, ODUflex won't address all Sten's points. It can handle mapping various sized layer 1 payloads into OTN well, but it can't aggregated subrate traffic as this needs layer 2 processing - e.g. Mobile backhaul where the physical connection to a cell tower is GbE but the data is only 200/300/400 Mbit/s etc. To economically tranport this, the aggregation point needs to aggregate this traffic efficiently, at layer 2. Transporting a lot of Ethernet traffic just at Layer 1 where each GbE is only 20/30/40% full is not an efficient use of network resources. Also, depending on services etc the node may also need access to Layer2 OAM data and ODUflex won't help here either. If you use OTN then you still need to perform layer 2 functionality and this means demuxing the OTN, processing at layer 2 and then remuxing a new OTN signal, which looses any end to end benefits that OTN might have added.

But ODUflex could be useful for mapping some non-Ethernet traffic and Sten's point is that the non-Ethernet traffic can be carried using TDM (possibly OTN) on particular wavelengths, as it already is very well today. Then the new Ethernet traffic that requires layer 2 aggregation/switching/OAM processing etc (which is where most of the traffic growth in networks will be) can be carried using techniques such as the Native Packet Optical approach on other wavelengths. WDM supports this separation of traffic very well and we don't need to force all traffic into OTN, especially within a metro enviroment.

As I mentioned in another post a few moments ago, once the relevant pipe has been aggregated enough to get it full enough and the traffic is all going to the same location then futher layer 2 isn't necessary within the transport network and OTN (or other approaches) could be used to transport this data - such as in the core of a network, where OTN does make more sense.

OTN will no doubt be very widely used in core networks where the main function is the transportation of full pipes. But in our opinion it is wrong to force all metro networks to use OTN everywhere as there is a very different set of requirements, it's not just smaller geography. I'm sure some metros will use OTN, but there are many live deployments now and other scenarios/deployments in the future where they won't use OTN and that was the main point.


Hope that helps clarify things as there were a lot of ideas to get across in a short period of time at the event!


Jon (Transmode - in case that wasn't obvious)

allanptj 12/5/2012 | 4:49:17 PM
re: Is OTN Overkill? Funny, I didn't stay engaged, I walked out on the tirade.
Pete Baldwin 12/5/2012 | 4:49:16 PM
re: Is OTN Overkill?

Chomsky -- It was indeed microseconds. Sten's slides don't have the particulars on that measurement; i think the main point was that they'd gotten latency way way down there in general.

Pete Baldwin 12/5/2012 | 4:49:16 PM
re: Is OTN Overkill?

> Funny, I didn't stay engaged, I walked out on the tirade.

Ha! :) 

sam masud 12/5/2012 | 4:49:16 PM
re: Is OTN Overkill?

You mentioned latency of "1.9 microsecond" Is that correct, or did you mean milliseconds? 

Page 1 / 2   >   >>
Sign In