Optical Feels the Facebook Effect

Facebook just put a face on a very significant trend.

We have been hearing for some time about how the web-scale companies -- Internet content providers, social media giants and cloud service providers -- are becoming bigger and more frequent buyers of optical networking technology. Though in many cases those customers didn't want to be identified in earnings reports or in news of contract awards, we could pretty much guess who they were, or at least get pretty close.

Finally, this week, one of them did appear by name in a vendor press release. So, why is it such a big deal that Facebook consented to be named by Infinera Corp. (Nasdaq: INFN) as a customer? (See Infinera's Big Customer Reveal: Facebook.)

Infinera's announcement didn't offer a tremendous amount of detail about the project, but did refer to it as "the world's longest terrestrial optical route" without regeneration, at 3,998 km. Heavy Reading Senior Analyst Sterling Perrin speculated via email earlier this week that maybe Facebook wanted "the cache" that comes from being able to make such a claim.

That probably has something to do with Facebook's optical coming out party, but it's also fair to wonder if it means that Facebook and other web giants are ready to become more public players -- and influencers -- in the optical network sector.

Want to know more about the networking moves of web giants? It's sure to be a hot topic at Light Reading's second Big Telecom Event on June 9-10 in Chicago. Get yourself registered today or get left behind!

Just a few days before Infinera touted Facebook as a customer, Facebook Network Hardware Engineer Yuval Bachar appeared at OFC 2015, offering his opinions on a number of optical market topics, from the need to more consolidation to the sluggish pace of standards development. (See Facebook Wants to See Optical Sector M&A.)

We talked about Facebook's desire for sector consolidation in our previous story, but here's what Bachar had to say about some other topics:

    On Facebook's bandwidth upgrade cycle for its data center networks: "Our data centers are growing by two-times to three-times their size every six to 10 months. We need a bandwidth upgrade every 12 to 18 months, so 40G is not working for us anymore, and we are looking at silicon photonics to take us to 100G, 400G and to 1 Terabit."

    On Facebook's view of fiber and module economics: "We are committed to making the transition from multi-mode fiber to single-mode duplex because it will be less expensive. Silicon photonics must be manufactured to scale. But, we trying create a direct channel to buy from the optical module manufacturers, and cut the middleman out. We have vendors [six to eight of them, he noted earlier] we can do this with, and the cost gets down to where we want it, about $1 per 1 Gbit/s."

    On pluggable optical modules: "Pluggable modules one day will not be viable. They are expensive and there are the problems of faceplate density and dissipation of heat. We are trying to drive down the cost [of pluggable modules], and this is why we like optics integrated onto the board."

    On industry standards: "Standardization in the world today is really slow, so we adopt a new technology, create an ecosystem for it, and then go to the standards bodies. We don't have time to wait for a standard to be adopted first, so we try to adopt the technology we think is best. We want to create inflection points for technology, but we want to be sure those inflection points also help create good business models for others across the industry."

It was clear long before the week's announcement and long before OFC last week that Facebook has an increasingly active voice in networking. That much has been clear from its involvement in the Open Compute Project and development of the open Wedge switch. (See Facebook in Production Testing of Open 'Wedge' Switch.)

But, in recent days, Facebook's voice on networking issues has been growing louder, and its identity as one of the top influencers of networking technology developments becoming ever more certain. This is a trend that both vendors and other big service providers need to understand. Face it.

— Dan O'Shea, Managing Editor, Light Reading

Page 1 / 2   >   >>
Nighthawk 4/8/2015 | 8:39:14 AM
Re: IFN without regen? hmmmm... I agree. The definition of regeneration can be of question but no doubt they can do this with only 1R "regeneration" or ILA if you'd prefer. Probably some gain equalizers used as well but they will not need any O-E-O on this distances.

Main issue as you say could be dispersion compensation but should be doable with last generation 100G coherent optics. Furthermore the BPSK has a theoretical distance of 5000km without O-E-O regeneration so questions if the 8Tbit/s claim is together with distance claim or separated. QPSK for 4000km can be of doubt.

freak0815 4/7/2015 | 1:53:25 PM
Re: IFN without regen? hmmmm... Again,

there are 1R, 2R, 3R regen and in INFN terms 4R regen.

1R - reamplification

2R - reamplification & reshaping

3R - reamplification & reshaping & retiming

4R - Digital Regeneration / Digital ROADM


You're absolutely right - this distance cannot work without any kind of regeneration and even dispersion compensation. So if they say there's no regeneration they mean it's no 4R regeneration. But there must still be some kind of 1R regeneration of course. If you don't consider amplification as regeneration it could be well without regeneration.

Like I mentioned earlier, with their Optical Express or FlexILS technology they can build 1R based links over such a distance easily.

Remember, this is a very cost efficient solution as there's no need for any intermediate line modules and latency should be even better without 4R.

tojofay 4/7/2015 | 12:38:58 PM
Re: IFN without regen? hmmmm... Take it up with INFN S'boat: Sunnyvale, Calif. – March 31, 2015 – Infinera announced that Facebook deployed an Infinera® Intelligent Transport Network  to light the world's longest terrestrial optical network route capable of delivering up to eight terabits per second (Tb/s) of data transmission capacity. The new route spans 3,998 kilometers and is deployed without any regeneration.

From INFN's release. Be an active shareholder and call them out.

sailboat0 4/7/2015 | 11:27:52 AM
Re: IFN without regen? hmmmm... Freak, 

you make my point precisely.  They are NOT doing it without regens.  They are simply regenerating using a different technique.  Digital regeneration at every node.  

What I meant by a point to point ULH network was a network with NO mid points and NO regens.  In this case, the article is a bit misleading.  IFN IS REGENERATING!  Just doing it digitally.

Now, if someone can prove me wrong.... go for it!  Show me a 4000km point to point with NO REGENS..... does not exist.  laws of propagation (physics) over single mode fiber.  By NO regen's, I mean absolutely NO regeneration..... including digitally!  

IFN indeed makes nice products.  and IF you buy sufficient number of line cards to have spare capacity, you can indeed turn up that spare capacity on deman and automated.  But, there is no free lunch.  you have to have those line cards installed.  (yes, IFN's line cards support multiple streams or links in one line card.... but you still gotta have them installed... which means paying for capacity you are not using in order to have that spare capacity in place to turn things up "on demand".)

I am an IFN stockholder.... so cool that they have Face as a customer.  

My comments were about the fact they are ARE regenerating, just not using traditional methods.  Doing it digitally at every node.  

My issue was with a misleading article with out sufficient detail to bolster the claims.  

check out the fallacy types listed in Rules For Writers, chapter 7.

freak0815 4/6/2015 | 2:59:56 PM
Re: IFN without regen? hmmmm... Sure Infinera can build optical point to point ULH networks. Already with the traditional line system they are using manual patches (call Optical Express) in the digital regen sites to build optical links spanning multiple digital links and with their new line system (FlexILS) this can be done even automated.

And believe me, with this distance, dispersion can be an issue - even when using coherent transmission.

Most notable: A capacity of 8 TBit/s over 4000 km point to point means QPSK should work over that big distance!
danielcawrey 4/5/2015 | 3:49:48 PM
Re: INFN understands... I'm happy that Facebook is getting involed in networking. Let's be realistic here: Facebook is now a big technology companie. Their scope is not just social media anymore – look at the Oculus acquisition. 

Facebook wants to push networking technology as much as it can – that is to its benefit. I don't expect that to cease. 
sailboat0 4/5/2015 | 12:09:20 PM
IFN without regen? hmmmm... Hello.

It is interesting that IFN claims this "world record" without regeneration.  While IFN has a good system / platform, they have not been able to break the laws of physics.  

It is a bit misleading to claim "no regeneration".  Yes, they are probably not using the typical "regeneration" used in most long haul networks: EDFAs + dispersion compensation for instance.  Or RAMAN.  But to suggest that they are able to reach this distance with out true regeneration is a bit misleading.  How about some investigative reporting into how this is actually done?  a span schematic would be quite helpful.  

serious doubts that this is not a single point to point network with NO adds or drops in the span  And confident that INF has not been able to defeat the need for regens.  They just do it differently.  At every node, IFN takes everything into the electrical / digital domain. The do processing, retiming, clean up of the signal etc. digitally in their asic, then re-convert the electronic signal back to optical.  So yes, maybe without traditional regen, but they are still regenerating the signal.  Just in a IFN way.

Now, if they have been able to break the laws of physics regards propagation of the signal down single mode long haul fiber, then I would like to see an in depth artical showing us how this Nobel Prize in physics caliber breakthrough actually works.  hmmm? anyone really up for that challenge?  

Great results for IFN, and happy they have Facebook as a customer.  But breaking the laws of physics?  I am ready for some one to explain this to me with real technical details.  Not marketing hype.  


jabailo 4/4/2015 | 2:17:28 PM
FB as primary content provider Facebook has been instructing its News Feed providers that rather than just linking to blog content, they should be posting directly to Facebook...including video.   That means Facebook will not just be text messages with hyperlinks, but an actually heavy duty carrier of massive content.   That means people will be relying on Facebook itself, not YouTube or news sites, for hosting and providing the back end throughput for video streams.   This deal may be serving notice to that effect.

tojofay 4/4/2015 | 2:12:36 PM
Re: More activity But isn't FB relying on INFN? Did they not just deploy INFN? I'd be willing to bet INFN developed their 2 new PICS for FB. The only "shelf" large scale photonic integrated circuits are coming off of is Infinera's and they are flying off of it!
wanlord 4/3/2015 | 10:09:35 PM
Re: More activity Others are slow to catch on. Facebook's OCP culture is to take risk, be bold and think big. Telcos and Cloud providers prefer to take it slow, rely on vendors instead of driving the technology. There are telecoms out there with much larger networks and data centers, but spend billions forking it over to vendors because the culture is to have someone else to blame instead of taking it on themselves.  They focus on hiring a bunch of cogs in the wheels to manage the vendors, instead of building a company with the smartest engineers. It's also not just about cost, it's about control of your own network, automation, tooling, and not having to go to the vendor for every feature request that takes months, instead they turn it around in days...

Page 1 / 2   >   >>
Sign In