Optics Evolution and Challenge for Router Interface

good morning everybody so so far this meeting we talked a lot of beats bytes packet and a lot of policy generally it's all focused on the software side right so next 20-30 minutes I want to change the subject a little bit touch base on the hardware so specifically on the optics router optics switch the switch optics and hopefully I gave a whole picture kind of like snapshot so what is the landscape on the router side so I'm a product line manager so work on the router side and why do we really care about objects on a router I so it's part of like you know it's a part of like a gaben you know go with the router and the one thing interesting thing with no taste for last few years we we offer a solution to operator to the end carrier more and more cost actually go to the optics side and so we need to really think about how do we actually offer a over a better solution from the cost capex perspective and also we have a mall chance to see when we really need some specific optics on router they're not available and is technical challenge or the timings not matched so here's a few topics on the cover next 20 30 minutes so the router optics versus transfer optics people kind of like easy to long together from the optics perspective right so automatically long them together with the transport if they are important piece of the network equipment a touch base on the on the router optics evolution and then we maybe walkthrough for you case in the real field 400 gig optics with rather what are the real practical issues let me touch base on what's next so many people might personal experience I by mistake about optics on the networking equipment automatically think about ahead you know like we have a huge optical transport industry over there a lot of investment go over there right so but many of the transport system most of the focus most of the investment really focus on called WDM system the CDC Rotom long haul ultra long haul what are the basic performance what are the spectral efficiency so they're kind of like a saga know compared to from the router perspective what is the requirement for optics so this is the picture right so the red part obviously is the is a transport gear and the blue side are the routers so from the router perspective what are the major point we need to consider its density we are looking for small form-factor lower power lower cost right so most of the most of the applications there gray objects could be pluggable could be fixed they're not WDM so from architecture perspective the top one is a majority was the majority of the architecture right so we have a router either co-located within a short distance who is a transport gear so we use a gray optics hand it over and then convert it to WTF the color interface some specific newer applications start with merge right so mainly for Metro for rien structure maybe point a point configuration so we start to see a uniquely DWDM directly on router so new architecture but that's only part of the story right so that's data playing integration and the beyond that we still need to do a lot of other work and but architecture majority of atom is top one so when we talk about the router optics we actually have all different kind of Aurora application and each of them they have a different requirement so on my on the picture on the right side is the Colorado it's really sensitive we are really pushing for is very high density very small form-factor lower power thermal cooling then we move to the middle to the edge daughter and the requirements become different ISO may be balanced from like low data read ten gig 40 gig hunter gig verses for power outer start a push 400 gig and even beyond and let me go further on the table are a skater be router on access aggregation so we might have a different requirement by mainly focus on thinking how do we save fiber a longer distance reduced cost and when a router is getting closer and closer to their cell tower we might need a requirement for an extended temperature my city in the cabinet have a size limitation have an airflow with Hartmann different direction so on a start focus on the hundred gig right so this is the year 100 Gig mainly right become a mainstream so we started from Leslie beginning with this decade laughter's started out for 100 interface the very first one pluggable is a CFP every two year also move along to come out a newer one reduce the size maybe half the size right from CFP to share P to share before it's kind of interesting couple years ago to three years ago there's a pretty big argument in the industry so we move from share P to what's the next one should we bet on the set before should we pay down the cursory 28 smaller form factor lower power this is really from the router perspective right so which one better fit so clearly in the industry tries to be 28 had a majority of the market moving forward we can see the beyond hunter gig let's say 400 big 200 why our multiple hundred gig is already on the horizon so the first one the 400 gig the form factors they share p8 so separate form factors are roughly the size of a share p2 but obviously right so change the data read from a hundred gig to quadruple to 400 big button forward I caught a second generation 400 gig what will be the fun factor a little bit day job right so in next year or two we have a few choices practical choice right in front of us one is a curious eye PDD double density and the other one is always have people so they both have advantage and disadvantage so always happy the form factors slightly bigger like twenty two millimeter which and so from overall density perspective from router side it's it's on par but you know a little bigger than the Kurds BDD but they can host higher power the curse BDD on the other side yes we have a more technical challenge to fit everything into that double density module but advantage for that is we can actually start from today to develop a long hard this lying hard with a host QSP form factor we can support case of twenty eight we can backward support care to be plus thirty big fall bite hanging moving forward we can support two hundred gig 400-year curtsy BDD but for OSF P is a brand new development so every single development on the router a lying heart causing millions dollars right so something interesting will come for next video next twelve months to 24 months we'll see which one ended up with a dominate the market then overall we talked about the router optics what are the challenges so from the top level we can see the major building block for Aurora Hardware one is the network processor that's it ASIC right either a commercial off-the-shelf one or all the major router company we have our in-house developer easy so every few years more like I don't say exactly follow Moore's law but kind of like a follow that path we increase the capacity Reilly increased density and so eventually we hit the point the Asics are already available we can support a much bigger capacity but the bottom picture just try to to show we are the use up all the front panel just mechanical space optical module all have size have a power right so we can only fit so much so this is the one of the challenge so ASIC development our page optical module bandwidth and we also constantly have the assignment challenge should we stay with the current for example this year 100 Giga module all we should move to the next higher speed module so you'll first jump into the high speed module obviously you don't have the cost benefit I have some other benefit but it's extremely difficult to see like early days 100 gig right so the the baseline hundred gig module cost need to be absolutely smaller than 10 x 10 gig but the early days you cannot achieve that by the real goal is not just 10 times it should be maybe 5 high maybe 4 times right we can improve the cost so same thing moving forward from 100 to 400 so physical size power and a router like so had a break time discussion yesterday so a kind of like a router total power increased so dramatically for last 15 20 years right so a one-car router right with full service might we're talking about 20 even 30 thousand Watts so many of them actually contributed by by the optical module right with those kind of very big power box how do we similarly manage them can we similarly manage them just for the optical module and then we have some other more like it I caught a trilling problem right so it's not a cutting edge but a lot of deployment over there that is rather aggregation router access router so maybe 10 gig but we want to save fiber right the Metro distance fiber 80 kilometer 40 kilometer can we squeeze more to the fiber they say bi-directional it's one way and they try to minimize the cost not a use WDM can we go a little longer distance even beyond 80 kilometer so there's many different kind of a challenge and another one is nowadays the router getting closer and closer to the self harbor right we have a internal kind of like survey it's double digit percentage of the smaller router actually need to sit in the field maybe on the curve maybe somewhere you know outdoor in it happening so we need extended temperature module so all those things are top level why we're facing some challenge to offer a good solution on the optics time for router let's back to under 100 Gig side again so this is also a kind of like this is more like the whole landscape most of them are cures have returned here yeah we saw it maybe 10 15 years ago for tanki have so many different flavors right but this is for me at least right so it's the first time ever for one optical form factor there was so many diversified different kind of flavor what are the reasons so the major reason is linked to a solution cost so mainly RightScale company and some big carriers we are considering the push optimize the design with different kind of performance diff kind of distance obviously associated with different hundred cost so you instead about a traditional define I'll are the ten kilometer dial r4 and the other end is isar so some other things in the middle start a pop-up CWDM for only support two kilometres with a lower cost possibly lower power PSM for 500 meter so many company think about link the router ice pine leaf sweet and innocent her outer 500 meter is a magic number so these PS m4 can i address that with lower cost and then we need a longer distance beyond 10 kilometer we need ultra long distance once back I don't want to call okay there is a transport year marks T marks converted to you know amplifier you know WDM transponder how do we address that distance so this is a kind of like snaps a picture for the hundred gig qurtubi 28 and some of them share before so let's quickly walk through some of the challenge in the field so from the early generation share few years back 2010-2011 and then the second generation share p2 all the solder each multimode fiber is pending Tang gig per Lane and now when we move to higher density of smaller form factor lower power extremely difficult to integrate all those lasers all those receivers into the smaller margin so now we use 25 giggling 25 Keegan RZ for laser for receiver right so I have this one customer big here on customer they want to know how do we address that issue shot which is the way to go I already have my all the multimode fiber in the field already connected I really like the higher density low cost new Narada how can you help me to link them together from 10 gig per Lane 25 gig plate one practical challenge may be a niche market but there's one challenge we haven't even sorted out what is best way to do it yet but one option is some company working on a mechanical adapter plug in one end on the CP 2 and the other end you can just around the curiosity 28 but some other next level detail the new cursor the 28s are you need a VC that also need to be addressed so the second one is the ER decent H a predefined standard 30-kilometer the whole industry really work hard from first generation CFP move the second generation share p2 we can support yahwah unfortunately for the new form factor to share before curitiba 28 at least up to today there's a technical challenge nobody can do it yeah it's you know they try the all this different way to look to increase the laser output power swap the the receiver with APD still cannot do it so you need to quickly put together a non-standard design might say right so it's called a 4wd em so the idea is we add FEC on a host board and to help to extend the distance without I if you see on a host port just this module optical module alone it can go 20 kilometer maybe 25 kilometer ish para one can also enter into operate with the in the field share p2 or share par-4 but where is limited distance if we need to go all the way to cover the 40 kilometer there's a challenge even this idea right so the modules not exist so I have been from the router from switch company's perspective we keep pushing the industry pushing the community on the transceiver side to develop that module for us I believe is coming within this year so the other one is yes further longer distance like 80 kilometer we start to offer you can see in the industry use a simplified design current hundred gig color share PCR to address at 80 kilometer but moving forward there's a CFP phone factor I so nowadays the the router interface already moved to catch twenty a year before how do we increase the density so the new one potentially there's two options one is already kind of like the coming at the door a security be 28 Zr but unfortunately it's not a foreign technology base it's painful technology based the other one cannot interpret we see exist one in the other end the other option potentially share p2p Co simplify design can also address this issue and the other one I used to use a story to explain I so it's like the carrier your facility is already in the field I it's extremely difficult to move your facility to shorten the distance and so one we have more and more router move closer to the field I to the curb to the to the cell tower they already have the pressure the optical module need to survive from negative 40 degree to very high temperature again this is another module is not agrees for the whole industry i've nowadays in the field the deployment they are all tanking and how do we come out something can address that issue so allowed the smaller one are you - are you smaller access aggregation smaller router and the backhaul order to increase the data read especially the upper link right from dengue 200 P there's another challenge so I believe the module could be available later this year wanna touch base a little bit on a 10 gig side right so even though industry is quickly moving to 200 gig but still there there's some requirement in the field based on hanging right how do we support a bi-directional beyond 14 kilometer it's the 80 kilometer 60 kilometer or even beyond we need some a special optical design how do we address more than 100 kilometer auto long distance when spent without a WDM system so one thing we noticed in last or whatever years right most of the network engineers they want to focus on a package side don't use the CLI to configure everything to link the rather than the switch instead of handing over to a transport and then there's a different domain so those are the some of the challenge and the other one is even we quickly move the router switch interface 200 gig we still have a lot of it hanging interface in the field how do we support higher density all the hundred giggle equivalent density tanking to address the dangling link there is a lot of other smaller are those are the challenge for tangling then we talk about the other way or arm moving forward already some big companies start to request one should we support young hunger geek 200-400 geek and be hung is there a market for order to support 100 Gig density higher than cure to be 28 Kurt Abbott 28 is like the default right the best so far so one way to think about we pump out all the data lane 200 gig we can support I said people as a woman to smaller form factor but that electrical interface hundred e electrical interface gonna take a long time to mature maybe another four five years maybe even beyond the how do we address the issue my next generation water you know sewage from the asic perspective we can support much higher density but from the interface perspective QSP 20s is not enough to take advantage of the the capacity for energy so another way is developed like I call the end by Henry to Mahendra gig for 100 Gig different form factor it's a little a chart right so that's theoretically different kind of a combination what we can do for n by Henry gig now we need to touch base a little bit on a 400-year from the ASIC perspective it's ready it's ready few years back that's one of the typical story when ASIC already can support a higher capacity unfortunately the optical industry they are trilling there's no 400 gig interface before so this is the first a year we're gonna see 400 gig pluggable module share p8 can be used and next year too as I briefly mentioned right so OSF P cursor BDD based foreigner gig will come to the market so there are benefit over there also challenge right benefits mainly lie on they say simplify the fiber management on the router each hunter gig come out fiber connected by Phi O'Hara so many hundred gig like the old days tanking even the fiber management is a challenge and the other one from package perspective right we can ease the concern for the leg bundle because we quadrupled hitter read but on the other side the early 400 gig might not achieve the goal from the cost perspective less than ten times of tangly this is the last topic on a touch base before the closing it's on the integral DWDM so the picture shows everything I for specific application Metro distance point a point we instructor we can simply find from the data plane integration from the top picture you honey door agree optics think about 100 Gig more expensive more 400 gig coming to the bottom of one you move the color transboundary directly on rather you read remove for this one simple point point link for expensive hundred gig module you also simply find the optical node in the middle because you don't need a host trans boundary you don't need a host plug ball module right and can be a design optimized design for this application much simpler may be amplifier you know filter something so it's the one way to think about on the other side when we have the integrate ability I'm our router we have the from the packet layer from the router perspective we have the full visibility what's going on on the transport side we can easily provision the service again by CLI line up another path hundred gig we can also quickly revolve triggered I far because we know what's going on under on a physically so there are some benefit but also a lot of limitation by the transponder the color transponder density is not as good as agria optics pluggable optics so there's a density mismatch and the product life cycle as a router typically it's a shorter they say may be roughly give or take maybe half the life cycle of the transport gear but nevertheless for some special application this is the one way to vote so let's briefly talk about what are the trend moving forward router optics so 100 Gig quickly become the new technique what I mean over here is 10 gig stay in the industry for a long time 15 20 years higher gig pave the way right stay long time ASIC and optics really dictate the cost of router it's your major network building block router is getting closer to the field to the wireless equipment so we need a smaller size less power maybe nc' complying extended temperature we used to call the client optics and lie optics right a WDM client I didn't learn if you open up some 100 Gig module inside the CWDM but it's a filter it's about apple wavelength and the last point over here is we can see in next few years the shorter reach less than one spent up the DN a-- coming because data center I so grow so fast 100 Gig go higher density you still want to squeeze him all channels into one five maybe only tank lamb in the 40 kilometer that's the 80 kilometer but your sail fiber that's the end of talk of an open for questions hey Lane Wiggly Cisco thanks for a good great talk can you talk a bit more about your view on breakouts in 480 break hard on 480 yeah so there are few challenge over there right so one is when we recount lesson usability I say example right so we can support the the industry can support a breakout fall by 100 Gig but that 100 Gig very possibly we need a hundred giggling time for based optics it fit in with the Q as a p form factor but not the same as the current available currently so we need a new hunter gig module to interoperate with the other side my memory card hi Samuel James Taylor Bluebird Network we've got legacy alcatel-lucent gear in our network and and I'm excited to hear about Nokia's plans are you guys moving forward are you integrating the ALU gear with legacy Nokia gear or how are you doing that on your roadmap to 400k so nokia have a multiple divisions ISO the router side actually come from mocktail listen somewhere under the same roof the same division router side and the transport side so we work very closely with my my counterpart on the transport side to start to address what are the next one the 400 gig as I said right so from the time to market consideration CRP 8 is the first one because that's the only available one this year and we need to build up some solution by to support 400 gig end to end but that's that's best plan all right thank you very much [Applause]

Channel: NANOG

Views: 3,671

Rating: 5 out of 5

Keywords: NANOG, NANOG 69

Id: kTgcOLmK1AY

Length: 35min 31sec (2131 seconds)

Published: Wed Feb 08 2017