WDM-PON Gets Its Day

4:50 PM -- As WDM-PON continues inching toward real commercial potential, it's gotten enough fans to warrant its own workshop. Or, so organizers hope.

The first WDM-PON Workshop will debut on Sunday, Sept. 12, at the 2010 FTTH Conference and Expo in Las Vegas. It's a full-day event promising papers, panels, and PowerPoints. [Ed. note: Woohoo!]

Organizers for the workshop include Bernd Hesse, a senior business development director at LG-Ericsson Co. Ltd. who got there through the Nortel Networks Ltd. acquisition of WDM-PON player Novera Optics. (See Ericsson Snaps Up LG-Nortel Stake.)

It does seem like a reasonable time to launch this kind of program. "I think it's key to show that, besides EPON and GPON, WDM-PON is a future-proof technology to consider, since the bandwidth demand is increasing," Hesse writes in an email to Light Reading.

WDM-PON is still several steps from reality, by most accounts, stalled by economics and the pending rise of 10-Gbit/s PON options. But the technology is available -- we're not talking about invisible hovercars here -- and it's getting advocates like ADVA Optical Networking . (See ADVA Opens Center, Wins Deal.)

There's a call for papers with a deadline of July 28. You can read more at Fiberin.com, an optical community Website created by Hesse and a partner. Deadline is July 28.

— Craig Matsumoto, West Coast Editor, Light Reading

astrithulaj 12/5/2012 | 5:14:53 PM
re: WDM-PON Gets Its Day <div id="autotrans" style="display: block;">
<h3 id="headingtext">

Dear all,

First let me apologize to you if you bother, but I am a master student at college and I need a software that enables me to do simulations for FTTH optical network. If you can help me in this regard had been very grateful to.


paolo.franzoi 12/5/2012 | 5:14:52 PM
re: WDM-PON Gets Its Day

What kind of simulation are you talking about?&nbsp; Protocol simulation?&nbsp; Optical simulation?&nbsp;



paolo.franzoi 12/5/2012 | 4:29:33 PM
re: WDM-PON Gets Its Day



The technology for WDM-PON is very much like the technology for Hovercars, available but so expensive that nobody will deploy it en masse for the forseeable future.

Want to judge it from a presentation?&nbsp; Ask the following question:&nbsp; When will your ONTs reach a sales price of $100?



Pete Baldwin 12/5/2012 | 4:29:33 PM
re: WDM-PON Gets Its Day

OK, I'm now told that WDM-PON workshops got held at ECOC about three years ago. As if&nbsp; anybody cared back then...&nbsp; :)

Duh! 12/5/2012 | 4:29:27 PM
re: WDM-PON Gets Its Day

...and OFC as well.

All of which misses the real issue with WDM PON, at least in the one (or two) wavelength per subscriber models:&nbsp; it ignores nearly 40 years of research and experience in packet switched networks.&nbsp; The central tenet of packet switching is&nbsp; statistical sharing of common infrastructure, resulting in a net gain in engineered system capacity. &nbsp; WDM PONs that isolate subscriber traffic onto physical channels eliminate any opportunity for statistical gain in the access network.&nbsp; Any period of time when a subscriber is neither transmitting nor receiving is a period when the channel is effectively being wasted, even if another subscriber on the same PON has packets enqueued to be sent or received. &nbsp;

Also,&nbsp; the economic design trade-off is presently between a 2.4/1.2 Gb/s TDM/TDMA PON with a 1:32 split and a 100 Mb/s WDM PON.&nbsp; Although the latter is more expensive, the former offers lower edge-to-edge packet delay (noting that in the upstream, this depends on reasonably good Dynamic Bandwidth Allocation policies).&nbsp; It also allows greater flexibility for service differentiation and pricing tiers.

Note again that this applies only to WDM PONS that assign a bidirectional channel (1 or 2 wavelengths) per subscriber.&nbsp;&nbsp; Architectures similar to that of the PIEMAN project, where WDM is used in the metro but feeder and distribution are TDM/TDMA, make good sense.&nbsp; There is also an argument for an architecture where the access network is colorless (i.e., splitter vs AWG at the remote node), but ONTs are tunable as well as having TDM/TDMA functionality.&nbsp; This depends on the cost of tunable filters and lasers relative to the cost of higher rate transceivers.&nbsp;

And as Seven points out, GPON and GE-PON have built up huge industry volumes, and driven down material cost and pricing.&nbsp;&nbsp; WDM PON is going to have a hard time catching up.

Duh! 12/5/2012 | 4:29:26 PM
re: WDM-PON Gets Its Day


I have tried to maintain a modicum of anonymity on this board.&nbsp; Suffice to say, you and I were competitors at one time, and it is entirely likely that we've met (or at least were in the same room)&nbsp; in real life.&nbsp; But on the Internet, nobody knows you're a dog.&nbsp; Anyway, I think a lot about this stuff as well, and it's good to see that we're thinking alike.

The wavelength agile TDM/WDM PON research is not just a university thing.&nbsp; Aegis Lightwave did a lot of work on the downstream piece of it a few years back, and published a couple of papers on a 4 wavelength tunable DWDM filter.&nbsp;&nbsp; The argument for that vs XG-PON was that the filter could probably come in for less than the incremental cost of a 10G-capable APD and receiver (relative to the 2.5G-capable APD and receiver).&nbsp;&nbsp; There was also an argument that OLT transmitters could be populated on an as-needed basis, and that there was some opportunity for grooming and traffic engineering.

On your last point, it is fairly well understood that in network engineering, the law of large numbers is your friend.&nbsp; Operators get away with more aggressive traffic engineering factors in the core than in the metro, and in the metro than in the access because larger host populations give rise to smoother traffic than smaller ones.&nbsp;&nbsp; Operators pretty well understand how much backhaul capacity they need to provision for a given subscriber population to achieve objective busy hour queue depths. &nbsp; On most current PONs, traffic engineering factor is limited by optical power budget and service bit rate, and is pretty close to unity.&nbsp; However, there's no reason why a 32:1 split GPON couldn't be used to offer, say, 200 Mb/s service.&nbsp; Note, incidentally, that DOCSIS 2.0 networks were often provisioned for trafic engineering factors of more than 100 with 1000 HP splits, albeit at considerable cost in busy hour latency.&nbsp;

paolo.franzoi 12/5/2012 | 4:29:26 PM
re: WDM-PON Gets Its Day



Given my history with FiOS and FTTH, you can imagine I have spent a very long time thinking about this stuff and I think what you have written is right on.&nbsp; I really found the 100Mb/s WDM PON stuff relatively useless (is it still owned by Nortel or did it get bought in all of this hubbub).&nbsp; A 1 Gb/s version might be more useful, but seriously if it is just to save fiber count in the ground then people are wasting their time.

I have seen some university thinking about Agile wavelength systems that allowed the sharing of many many customers around a single high bandwidth, fast tunable laser.&nbsp; Is that a possibility?&nbsp; Maybe.&nbsp; Having reflective or regenerative ONTs is for further research to get their costs down dramatically.

But let me leave you all with this thinking.&nbsp; I know that Verizon FiOS started with a uplink to the OLT of 1 OC3c per 50 PONs.&nbsp; I know this for a fact.&nbsp; The next expansion of bandwidth to a FiOS OLT was 1 OC3c per BPON shelf or 20 PONs (max).&nbsp; This then went to 1 Gb/s per shelf on BPON and the capability is there to go to 2 Gb/s.&nbsp; I know that the GPON RFP wanted much bigger uplinks potentially but realistically I think they are doing similarly 1 - 2 Gb/s.&nbsp; So are we thinking they are going to do 10 Gb/s uplinks on the 10G PONs? Sheesh I watched them struggle for the costs of doing a router upgrade to add a GigE per shelf.&nbsp; So will they put 200 PONs on a 10GigE?&nbsp;



paolo.franzoi 12/5/2012 | 4:29:25 PM
re: WDM-PON Gets Its Day



Yeah, I remember Agile Lightwave.&nbsp; I have dealt with some university types that were trying to look to using 10G Fast Tunables.&nbsp; I just always sit here an ponder how is that going to be cheaper.&nbsp; I have had other interactions with folks working on high power, wide bandwidth sources.&nbsp; That would allow some filtering to be done to create all kinds of interesting possibilities.

Where I was headed with my last bit wasn't the oversubscription thing, which I very well understand.&nbsp; Really it is the idea that a 10G access solution really only carries value if you run it through a 10G aggregation then to some higher order metro and core solution.

Now, where I was going with that was the entire problem of 10G access ports on the aggregation routers (in the FiOS case E320s).&nbsp; I am just thinking about the cost of an E320 upgrade to 10G access ports to OLTs and 40G (maybe) uplinks to their metro.&nbsp; The cost of the router ports was astronomical and I was just thinking that the costs here are massive unless they aggregate the ports up even more than they do today.




Pete Baldwin 12/5/2012 | 4:23:59 PM
re: WDM-PON Gets Its Day

I'm at the WDM-PON conference now (yes, in Vegas on Sunday), and Zulfiker Morbi of Applied Optoelectronics Inc. has just referenced this very comment thread.

His presentation talked about lowering ONU cost by using injection-locked Fabry -Perot lasers.&nbsp; They've put these into SFP modules, and his assertion is that the costs can come down enough to 1.2 to 1.5 that of GPON.&nbsp; (We're talking solely about optical block costs here.)&nbsp;

10G GPON costs, by his count, would be 10x that of GPON in the asymmetric case, or 20-50x for symmetric.

His point was that it's possible for WDM-PON costs to come down. "This is not invisible hovercar technology," he said.

He added that the more important question than the $100 ONT, in his view, is the total network cost ... which, admittedly, is what you hear from most vendors.

I'm not saying I've vetted these numbers -- this came across the presentation screen minutes ago.&nbsp; It's interesting, though, and not just because Morbi put the&nbsp; LR page up on screen.&nbsp; (Thanks for the plug, btw, Zulfiker.)

Sign In