Doubts Still Dog IPoDWDM

Cisco Systems Inc. (Nasdaq: CSCO) is happy about a report saying IP over DWDM (IPoDWDM) is in a majority of carriers' plans, but analysts still aren't convinced the technology is all that widespread.

For a lot of those carriers, IPoDWDM is only being used in a few spots where it can cut costs, says Michael Howard, principal analyst of Infonetics Research Inc.

"There are a couple of big Asian operators that are using it today, but overall, it's pretty minimal," Howard says.

IPoDWDM involves putting a tunable DWDM interface on a router. The more normal practice is to have the router feed a transponder system, which in turn assigns the traffic to a DWDM wavelength. So, IPoDWDM can eliminate that transponder box, making the network a little simpler and, theoretically, a little cheaper to run. (See Cisco's CRS-1 Goes Optical.)

Cisco started pushing the technology about five years ago, and Alcatel-Lucent (NYSE: ALU) and Juniper Networks Inc. (NYSE: JNPR) now support it as well, albeit with less gusto. But when it comes to converging packet and optical layers, more attention has gone to packet-optical transport systems (P-OTS) and similar devices, where routing functions get added to a DWDM system, the philosophical opposite of IPoDWDM.

So, Cisco was thrilled by a recent Infonetics survey where half the operators polled said they were using IPoDWDM already or would be using it by the end of the year. Seventy percent said they'd be using it by 2012.

Cisco popped the champagne with a blog entry -- citing slightly different Infonetics numbers (85 percent adoption of IPoDWDM beyond 2012) -- about IPoDWDM finally "crossing the chasm," the overused industry phrase for a technology going mainstream.

Howard admits to being impressed at the breadth of carriers considering IPoDWDM. Quite a few "were pretty sure there would be a good reason to use IPoDWDM, in a cost-saving basis, by 2013," he says. But he adds that less than 10,000 IPoDWDM interfaces are in place. (Nearly all of them are Cisco's, by the way.)

But Cisco thinks it's unfair to call IPoDWDM a one-node fling, and the company denies that the technology is being used only opportunistically. "We have many service providers that use this broadly as their main backbone transport strategy and are now moving it into the metro," writes Mike Capuano, Cisco's director of service provider marketing.

Cisco isn't disclosing who's using how much IPoDWDM, but its named customers include Comcast Corp. (Nasdaq: CMCSA, CMCSK), Deutsche Telekom AG (NYSE: DT), Free , KazakhTelecom , Netia Holdings SA , Sprint Corp. (NYSE: S), T-Com , and Telia Company .

Still, P-OTS and other forms of packet/optical convergence seem to be more on carriers' minds. "The more I talk to vendors and operators about that, there just seems to be less and less need" for IPoDWDM, and more interest in using P-OTS to bypass some core router ports, says Heavy Reading analyst Sterling Perrin. (See Vendors Target the Packet-Optical Core.)

"It's almost the most universal thing I come across. They want to offload the burden on those routers for transit traffic. Which is the opposite of IPoDWDM," Perrin says. "I was talking with Comcast Corp. (Nasdaq: CMCSA, CMCSK) last week, and they're interested in that." (Comcast was an early proponent of IPoDWDM.)

Another problem, especially for large carriers, is the operational question that IPoDWDM creates, because it blurs the definition of which part of the operation owns the wavelength. If it's the router side, then those people need a way to find out if the wavelength is even available on the optical transport side.

Control-plane innovations such as GMPLS are supposed to help with that issue, but in some cases, the answer comes down to a router person going over to the optical group and asking for a wavelength, Howard says. That's not particularly scalable.

Cisco notes, though, that this is a well-known problem that's fading away. "The siloed domains of 2005 have already merged in many service providers" and will exist in just 5 percent of operators by 2012, Capuano writes, quoting more Infonetics numbers.

What's interesting is that Cisco could hedge its bets by adding P-OTS. Its recent acquisition of CoreOptics appears to be motivated by the coming 100-Gbit/s optical generation, but Eve Griliches of ACG Research thinks it also points to a P-OTS plan. (See Cisco Renews Optical Focus With CoreOptics.)

Analyst Ed Zabitsky of ACI Research (similar name, completely different company) agreed in a note published early this week: "Their acquisition of Coreoptics shows their understanding that intelligent optics will subsume some data transport dollars from routers. That is the Cisco Way. Let customers bet on Cisco, not on a specific technology."

Nokia Networks is similarly playing both sides. It can support IPoDWDM, through a partnership with Juniper, but it's also developed an OTN switch. (See NSN Adds Packet-Optical Punch.)

— Craig Matsumoto, West Coast Editor, Light Reading

Page 1 / 2   >   >>
michaelinfonetics 12/5/2012 | 4:26:47 PM
re: Doubts Still Dog IPoDWDM

I believe that operators are moving toward IPoDWDM, even while others in the industry may still be in doubt. Our research clearly shows more and more operators plan to use IPoDWDM over time--70% by 2012 and 85% by some time after that, including 73% of the incumbents. Most of this IPoDWDM use is in the high volume Metro/Core parts of the network. Even though some of the largest operators today may only deploy small numbers of IPoDWDM interfaces, I believe that the future holds the deployment of more, not less IPoDWDM, and by more operators, and then most operators. Also, I would be careful to pit routers with IPoDWDM directly against P-OTS equipment, as IPoDWDM does not mean a direct competition with P-OTS equipment--P-OTS are quite complementary. In our research, we found that many operators see a much stronger role of WDM in the router over the next few years--to the extent that many of the P-OTS functions will be present on router platforms in 2-3 years. The roles of router, CE switch, P-OTS, and packet/WDM gear will more heavily overlap, and provider choice of product will be formfactor, placement in network (access, access/aggregation, aggregation, core), packet/video/TDM traffic mix, and TCO. Cisco does have many operator customers that have a single ("converged") operations staff, that buy Cisco routers/IPoDWDM along with 15454/ROADM as a solution that suits them well, and reduces their overall spend. In many cases, especially given the growing IP video traffic, the purchase of more router ports can mean a lower total network cost and lower operations cost as well. As the large incumbent operators converge their data and transport staffs into more aligned workgroups with common processses, it opens the door for the equipment they manage to be converged as well.

michaelinfonetics 12/5/2012 | 4:26:46 PM
re: Doubts Still Dog IPoDWDM

Good point, Sterling, and thanks for asking for the clarification--I'm sure other readers are curious. Because of the potential problem of different meanings for different folks, we were careful to define IPoDWDM as using colored optics on the router. Our definition: IP over DWDM is the integration of DWDM optics (colored optics) in a router, which connects the router with an optically switched DWDM layer.

We were careful to not let them forget the definition when we asked the key IPoDWDM question: For which of the following connection types are you likely to use colored optics (C-band or tunable optics) on routers (also known as IPoDWDM), and by when?

I hope this answers your question. I'm interested also in knowing what other readers think about the definition and future uses of IPoDWDM.


Sterling Perrin 12/5/2012 | 4:26:46 PM
re: Doubts Still Dog IPoDWDM

Hi Michael,

How did you define IPoDWDM to the operators for this study? (this is not a snide comment, just a question). The term can be used very broadly ("I have an IP layer and a DWDM layer, so I do IPoDWDM" or much more narrowly).

My comments in this piece were specific to the core network. One trend that we anticipate is integration of DWDM optics on Layer 2/3 switches - i.e., products that are being sold for Carrier Ethernet Transport applications. And this certainly makes sense. These are transport products, so the argument applies to collapse the transport layers into a single device (just as P-OTS does). This could be called IPoDWDM, but, to me, it is a different animal than the original IPoDWDM concept of combining the "full blown" router with the DWDM transport functions. One big reason is that the ports on the L2/3 switches are priced for transport and the ports on routers are priced to include the full routing that the boxes perform.

I'm curious what the LR universe has to say on this topic of what is IPoDWDM and where it will be applied.


ohub 12/5/2012 | 4:26:44 PM
re: Doubts Still Dog IPoDWDM

Good discussion! The definition of IPoWDM depends on whether we allow the WDM layer to have lightpath routing functionality (i.e., optical routing and switching in ROADM). With lightpath routing functionality supported, there is no issue of “they want to offload the burden on those routers for transit traffic.” as the transit traffic can be offloaded on intermediate routers through optical bypass at the lower-layer ROADMs. In this sense, IPoWDM is the most efficient architecture. However, if the definition of WDM layer only means dumb point-to-point WDM systems, then the issue related to transit traffic raises. IPoWDM does show strong disadvantages in handling the transit traffic.

somedumbPM 12/5/2012 | 4:26:41 PM
re: Doubts Still Dog IPoDWDM

I have been hearing about WDM ports on routers for a while, but I have only seen a few actually implemented and those were shooting Ethernet traffic over very short distances.

I have always wondered how amplification and automatic dispersion compensation are handled by the router ports or are these solutions solely meant for shorter connections which therefore do not have to address these type of issues. 

I assume that FICON and Fiber Channel were not able to be addressed in the past -at least not cost effectively.  They may be able to be addressed now with FCoE, but it is surely not cost effective (or trustworthy) now either.

Due to these perceived issues, I have not delved much further down this rabbit hole -feel free to prove me wrong.  I have no issue in being enlightened, I actually enjoy learning, but no marketing spludge please.

gazdowski 12/5/2012 | 4:26:40 PM
re: Doubts Still Dog IPoDWDM

Hope this extract from Alcatel-Lucent WhitePapers Library on CBT-Converged Backbone Transformation consitutes good answer to that article.


"With Internet traffic expected to grow by a factor of five between now and 2013, fueled by video on demand and mobile broadband, service providers do need a way to relieve the strain on their core network while keeping their investment costs to a minimum. Consolidation of the IP and optical layers is crucial to achieving operational efficiency and cost savings while providing end-to-end visibility, manageability and control. The Alcatel-Lucent Converged Backbone Transformation solution provides the lowest cost-per-bit delivery with continuous bandwidth scaling, automation and reporting regardless of service mix.

Economic benefits

Lower operational and maintenance costs

<li class="bullet first-child">Best of breed transport and IP platforms providing efficiencies in power, space, sparing, network planning and training </li>
<li class="bullet">Pre-integrated and pre-validated E2E solution including integrated management tools across IP and optical domains </li>
<li class="bullet last-child">IP transformation expertise including business consulting and multi-vendor maintenance services </li>

More efficient utilization of network resources &amp; assets

<li class="bullet first-child">Better router port and optical link utilization with 3 traffic grooming options (lambda, port, sub-port) to adapt to service mix </li>
<li class="bullet">Reduction in costly OEO conversions with transponderless optical interfaces and fast-track IP offload paths </li>
<li class="bullet last-child">Transparent in-shelf evolution from 10G to 40G and 100G </li>

Accelerated service provisioning, fault isolation &amp; troubleshooting at lower cost and complexity across IP and optical domains

<li class="bullet first-child">Zero Touch Photonics eliminates error-prone manual optical network configuration with remotely tunable transponders </li>
<li class="bullet last-child">Advanced optical layer management of each individual wavelength channel with Wavelength Tracker </li>

Driving least cost per bit with at least 30% CAPEX &amp; 50% power savings

Deployment, management, end-user benefits

Faster time-to-market

<li class="bullet first-child">E2E provisioning with integrated element &amp; network management tools </li>
<li class="bullet">Solution 100% pre-tested for end-to-end validation and interoperability of underlying elements </li>
<li class="bullet">Worldwide Network Transformation expertise gets it done right the first time </li>
<li class="bullet last-child">Cross-layer network planning &amp; design </li>

Simplified management

<li class="bullet first-child">Cross-domain visibility, control and fault management </li>
<li class="bullet">Wavelength Tracker:E2E power control, monitoring, tracking and fault-localization for each individual wavelength </li>
<li class="bullet last-child">Multi-vendor maintenance"</li>
CardinalX 12/5/2012 | 4:26:38 PM
re: Doubts Still Dog IPoDWDM

Is Lightreading trying to make it look like IPoDWDM v P-OTS? In real networks both will be needed. IPoDWDM when a router generates enough traffic to efficiently fill a wavelength and P-OTS when sub-wavelength grooming is required to efficiently pack wavelengths. Not all the traffic that P-OTS will groom comes from a single router network, it may be grooming traffic from legacy networks or other router networks.

Sterling Perrin 12/5/2012 | 4:26:37 PM
re: Doubts Still Dog IPoDWDM

CardinalX -

I see many instances in telecom where products/architectures are "not mutually exclusive" but they are still competitive. I get that there will be cases where IPoDWDM makes sense in the core, but, if P-OT is ultimately used through most of the core network, then I think that P-OT&nbsp;would be considered the winning architecture of the two.

Looking at things from the opposite view, if an operator builds most of its&nbsp;core network on IPoDWDM, then what is the need for a P-OT element in the core? It seems that optical elements would be used as ROADMs primarily, without need for packet functions on the optical transport box (this is what defines a box as "packet-optical" as opposed to just "optical"). And, if the transponders are located in the router, then they are not located in the optical box, so that capex spend shifts from the optical transport element to the router element. This too, seems competitive, not complementary.

IF you are saying that P-OT is to be used in the metro and IPoDWDM is to be used in the core, then I see that scenario as complementary&nbsp;in that Metro P-OT would be complementary to core IPoDWDM. But alot of suppliers are investing in beefing up with metro P-OT boxes for core applications. So, they definitely have the core in mind.


Pete Baldwin 12/5/2012 | 4:26:25 PM
re: Doubts Still Dog IPoDWDM

You might be right, but I think we're correct in pointing out that the technologies are competing against each other. There's still some debate over when either one is useful, as Sterling pointed out.&nbsp; The two are duking it out for positioning in the network.

I'm expecting Cisco to come up with a P-OTS product of its own, eventually.

Pete Baldwin 12/5/2012 | 4:26:25 PM
re: Doubts Still Dog IPoDWDM

&gt; I actually enjoy learning, but no marketing spludge please.

Anyone else amused by the response to this? Gazdowski must have a sense of humor.

Page 1 / 2   >   >>
Sign In