Ethernet and Carrier Networks
Date: Fri, 08/02/2013 - 15:12
Senior and executives from service providers and carrier solution vendors discuss the relatively recent role of Ethernet in bridging the “telechasm” separating the LAN and WAN
Tam Dell'Oro, Founder & President, Dell'Oro Group, at NetEvents Global Press & Analyst Summit
PHOTO / telecomkh.com
As Ethernet approaches the universal language predicted by Nan Chen, it looks less and less like a networking technology, and more and more like a service offering –with serius implications for the carrier market. The panel at NetEvents Global Summit, California, needs to identify the critical issues and lay down a new road map for the way forward.
Panellists: Howard Charney, Senior Vice President, Office of the President, Cisco; Said Ouissal, Vice President, Juniper; Mark Durrett, Director, Marketing, Overture; James Walker, Vice President, Managed Network Services, TATA Communications; Yali Liu, Executive Director, International Network Planning, Verizon
Introduced and chaired by Tam Dell'Oro, Founder & President, Dell'Oro Group
Okay, so panellists come on stage and I'll just start by giving you folks a little background. But before we go into the background, I just want us to take a reality check. All the people here that planted the seed, developed that seed of Ethernet do you realise what great jobs you've created for people all over the world, really it's fantastic, fantastic what you folks have developed.
Anyhow, I'm going to take about two or three minutes and I'm going to give you a little background about Ethernet and Carrier networks. Okay, just to give you a frame of reference or a little perspective on Ethernet and Carrier networks, Ethernet became popular in local area networks years ago. You had structured cabling, before structured cabling it was unstructured cabling. Do any of you guys remember that thick black coax cable above the ceiling rafters? There you are, good, good thanks I just remember the one coming down.
Anyhow -- and do any of you remember when Token Ring was bigger than Ethernet? Yes, so we had some proprietary structured Ethernet. What this shows us shared Ethernet and Token Ring we had FDDI, ATM. Does anybody remember Daisy systems? Alright, and Proteon? Great, Cabletron? Yes, yes.
Anyhow, Ethernet in enterprises it was quite popular in the 90's and then sometime around this early 90's timeframe we came to switched 10MG Ethernet and switched 10MG Ethernet only lasted a couple of quarters and then bam it went over to 100MG Ethernet and it's Howard's company and a couple of others that just turned the whole market on its ear. Look at the scale there it was rapidly up to a $10b market. Right now this is just Ethernet switches. There's Ethernet adapters, you've got Ethernet and optical transport, you've got a lot of other Ethernet but this is the mother of Ethernet here in the LAN and we are at about $20b a year.
Just to show you the longevity, now just to turn to the panel here and set you guys up, Ethernet started out in Carrier networks back in the early 2000 timeframe using the LAN type of equipment in the carrier networks. But then we got these Carrier style Ethernet switches and routers, and as you can see they started replacing those frame relay ATM switches. And this is a relatively new market and it's growing quite rapidly. New applications like LTE will probably drive it even further.
So now we are going to switch over to the panellists here and we are going to ask them to each talk about how Ethernet is important to their business or how it's been important. And perhaps, Howard, we'll start with you. Howard co-founded 3Com, and then started Grand Junction Networks which innovated 100G switching or 100MG switching. We've got Said over here at Juniper Networks, came from Redback, then went to Ericsson now over at Juniper on product and technology. We've got Mark over here who went to start [Hatteras] Networks and is now over at Overture. And we've got James here at TATA Communications who is spearheading the Ethernet business at TATA. And we've got [Yali] over at Verizon who is in product and technology on Ethernet and Internetworking, so please Howard.
Thank you, Tam. So you have to understand that Ethernet, and I know you do understand this, comes from the LAN world and it really wasn't a Carrier technology. And in fact we fought viciously wars to -- specifically about the physical layer, and I don't understand -- I do understand why we did that, but now it all seems for nought because you remember what John Shoch said earlier this morning, he said that you know we don't really have Carrier access anymore, and we don't really have multiple stations accessing the cable at the same time, and we don't really have collisions anymore, so what do we really have.
And it's a really -- it's an excellent observation because now despite the fact that we fought those physical wars we now only have what remains which is the frame. The frame remains what's really valuable. And even though we did fight those wars in the standards bodies, the frame is the heart of gold. And it is the frame which allows us actually to move forward and make it, make speed irrelevant because if we really were attached to the physical layer we couldn't be going to 100G and 400G and maybe a terabyte, it would just not -- it just wouldn't work.
So what's going on right now is we are making this Carrier if you will suitable by keeping the frame and by moving to different physical interfaces that serve our needs. I used to use this rule of thumb which was every time we changed the protocol you're going to take a 50% performance hit, every time. And so there is this beauty in not changing the frame. Now whether that rule of thumb is right or not is not what's important, what is important is that Carriers are in the business of selling services and making money and if the efficiency goes down of the network this is a bad scene, it can't work, so by retaining the frame format this is what's gold.
And so what's happened over the years is the practice has evolved. And how has it evolved? Well we used to -- we would send Ethernet over SDH and Sonet, or we would send Ethernet over IP, MPLS and now as you -- what the Metro Ethernet Forum stands for is we go to market with Carrier-based Ethernet, we don't have to send it over something else it's a service. And so this whole thing has evolved now so we don't have to change the frame. And that's really, really a good thing.
And so when you ask the question is this important to us, it's really important to us because it has become the go to market. So I asked our people is it a -- is this a criteria, a buying criteria? It turns out not it isn't. And you might say well how could it not be if it's that important to you? It's so important it's not a criterion it's to get in the door, its table stakes. If you don't have this capability, if you don't have this ability to provide these services you just don't get to play. So it has become so integral to our business that it is synonymous with our Carrier provisioning and sales.
Said or Mark do you want to take up after that?
Sure, I can go ahead. I think coming also back to the question that Bob asked earlier what does it mean to Juniper. I think [Ethernet] for us is an enabler, it's an enabler for our business. If you look at the other switching and routing business the majority of the ports we ship today are Ethernet ports. Now on the switching side you probably are not too surprised because switches we had the LAN wars in the 90's and the Ethernet came along and won. But I think if you look at the routing business when we as a company built basically the first purpose-built router for the Ethernet we didn't necessarily think about Ethernet as the primary means of connectivity, we built it for any interface Ethernet was just an interface that we put on our routers like we put Sonet SDH etc.
Now if you look at what happened in the early decade last decade in 2000/2003 I remember being with customers and every customer was asking us, talk to us about Ethernet, talk to us about what you're going to do with Ethernet. And in the beginning you go like well okay why are these, everybody is talking about Ethernet and in the old days we used to talk about Ethernet as just an interface and now the importance of Ethernet went up and we were like why? Well we think Ethernet can help us to keep up with bandwidth growth, we think it can make it cost effective for us. We think there is now an industry an ecosystem that has scale and that scale we think we can reuse in the service provider business.
So we went off, and actually for us the answer was to actually build the first Ethernet router, a router completely engineered for Ethernet and it meant it was Ethernet in the design process not at the end where like well okay the router is done what are we going to put around it. No, Ethernet was in the first step of the design process and we talked about -- we had to talk about the physical footprint, the Ethernet density, the price per port that we wanted to meet, and how we were going to be able to scale that up over time from a 1G to 10G to 100G today as a kind of our density leading design criteria.
So Ethernet is very important for our business, and we are very glad for all the work that has been done by a lot of people here in the last 40 years, and thank you for that.
Thanks Said, Mark?
Really I'd like to start where Said left off and talk about as you asked what does Ethernet mean to our business. Well my company Overture Networks is in the business of providing Carrier Ethernet hardware, software and services for service providers at the Metro Edge. And what does that mean? That's really that last mile of connectivity to businesses and to mobile towers and to data centres.
And so what's been fascinating for us and what's really -- well let me just back up. The reason Ethernet is important to us is because it's important to our customers. And really what Said said rings true with our customers in that the service providers had come to us and said we believe in Ethernet because we believe it's a way to deliver higher bandwidth at a lower cost per byte and we can build it on a reliable ecosystem that's evolved.
And what's been fascinating to watch is how the Ethernet has evolved in the Carrier space over these years. I remember the mid-90's when there was this thing called transparent land service, and hopefully some of you guys remember that, but it was really run out of a couple of cowboy operated service providers that really thought they could get a jump on some of the incumbents by delivering really high-speed connectivity in fibre rich areas in metropolitan cities.
And that evolved from small competitive operators into the major operators. I remember Bell South had a service they called [Nimley] and it was an Ethernet based service where they deployed an enterprise switch in the central office, and they put little [media] converters down at the edge of the network. This has really taken off obviously and evolved to the point where there is millions of connections around the world, and Ethernet as a service.
And it's amazing because it started off as dedicated point-to-point connections often Ethernet over Sonet, but it also simultaneously because a sort of dedicated Ethernet connection out to a building. But then as branch offices needed to come online with a corporate overriding switch to Ethernet service then technologies like Ethernet over voice-grade copper came onboard and so we have Ethernet over bonded copper now, Ethernet over T1's and E1's and Ethernet over the air, there is microwave Ethernet technology now.
And we really are at a place where the ether can be anything, and it's an amazing time. We now have Carrier Ethernet 2.0 which has really taken Ethernet to the next level in terms of classes of service, delivering multiple classes of service and manageable performance assured services, and then the ability for service providers to have a standard way to interconnect with each other around the world. So all this really drives the need for Ethernet in our business.
Thank you. James, why doesn't James start off please by saying how he's been in that business five years and its grown 15-fold?
Yes, so I've been around longer than that obviously but in TATA I've been with the company for five and half years and we've grown the Ethernet business by 15-fold which is relatively simple when it's sort of single-digit millions but it gets more interesting once you get to tens of millions and now we are hundreds of millions. So it's a very, very fast growing industry. As Mark said the reason why it's key to us is because it's key to our customers. And our customers are saying that Ethernet is the way that they want to do certain tasks, they don't want to everything over Ethernet in the WAN but they do want to do certain important tasks.
And I think the -- it's almost like the historical origins of Ethernet are coming back to support it once again. It started off in the LAN and in fact it's the importance of the LAN today that is making it so important in the WAN. It's so important in the LAN because it's what connects together data centre devices, so servers and all those sorts of things are all, even if you're doing fancy things at a higher level, you are doing things with VXLAN or something it's Layer 3 you are still fundamentally plugging in an RJ45 jack or whatever it might be into the back of the server. And that means that Ethernet starts all the way from the very starting point of the Cloud all the way through.
And if you can drive connectivity between those servers so that it stays within the same medium, within the same Ethernet frame all the way through from end-to-end then you have something that's really significant for the customer. And it comes back to a theme that's been quite frequently talked about this morning in particular, which is simplicity. The last thing the customers want to do is start laying multiple things on top of each other, and then they have competing interests from different protocols, Ethernet wants to behave in one way, IP wants to be routed in a different way and trying to get those two things to line up.
The fact that Ethernet can be used to extent data centres and make -- we talked about virtualisation in the previous session, virtualisation of data centres starts to get very interesting. If you are virtualising all your applications and all of your servers then let's also talk about virtualising your data centres so that Singapore and Hong Kong at least from an application perspective almost behaves the same way as if they are connected directly together on the LAN. That's quite transformational as far as the data centre designer is concerned.
And I suppose the final aspect that we see is that because Cloud has come on much faster than anybody expected it to do, it's grown much, much more quickly than anybody thought it would, particularly as those of us that remember it we all had a go at it with application service providers if you remember those? And in fact I would argue that really Cloud is a fancy version of mainframes really and shared access to mainframes. So we've seen this many times before, but now Cloud has suddenly really taken off in a way enabled by Ethernet and so on that's really exciting. As that happens and these environments become more and more large Ethernet is absolutely there from the very beginning. And extending that across the WAN becomes something that customers really are interested to do.
Thank you James. [Yali] want to talk about Ethernet with Verizon?
Yes, sure. So Verizon as a global provider that we definitely are there to provide whatever the services that our customers need, Ethernet is one of the fastest growing products we have seen in the last few years. And our customers require Ethernet in our view for I believe three reasons. One is that Ethernet provides the speed, the higher speed up to 10G in our case as a standard product. And number two it provides differentiated class of services, so it depends on the customer needs then the customer could choose different service levels and pay different prices. And number three is that the customer is able to increase the demand for bandwidth dynamically, so with the Ethernet the customer could chose to up their speed from 3MB to say 10MB or up to a 1G and without having to change the physical interface.
So with that we have seen double-digit growth in the last few years in [overall] Ethernet services growth in our product portfolio. And also from a global perspective actually specifically outside of the US we have seen very high adaptation of native Ethernet as the access to our customers although we are able to provide over 80% of the solutions outside of the US actually from a local access perspective as the Ethernet. And we work with hundreds of other service providers in addition to our own [natural] network to provide that to our customers.
Wonderful, wonderful. Does anyone out there have any questions for the panel here? Then I'll throw out another question to you folks. And if someone does have a question their hand and we'll tag you. Oh we've got two questions, okay. Mike right behind you.
Alan Weissberger - IEEE
I am Alan Weissberger with IEEE. I worked on the Ethernet First Mile Standard somewhere around 2000 to 2003, and it took maybe six years after that Standard was completed, 2009, for Carrier Ethernet, Business Ethernet, Metro Ethernet services to be deployed and then to ramp up. Why did it take so long? A lot of start-up companies including the one that I was in didn't have deep pockets we went belly-up in the [interim]. Why did it take so long? And what's going on now that causes that market to ramp up?
I mean I can -- we are in an interesting situation at TATA our home market or one of our home markets, we have India and South Africa as home markets, both of them have got historically relatively poor last mile infrastructure. And from our point of view when we looked at things like Ethernet and the First Mile it requires a copper infrastructure. If you look at India there are 40m copper lines in India that's it and it's declining. So this is for a population of around about 1.1b people. So that's almost unusable as a technology as far as we are concerned in our home market.
Now change that to look at an overall at a business we are quite unique that we have a home market but in fact three quarters of our revenue is generated internationally. So we are all about long haul intercontinental capacity, and we now with Ethernet services we reach 64 countries today with our own capability and then we extend beyond that through NNIs.
Every single market we go into has got different situations, some of them have got legacy service providers who have copper infrastructure and they have implemented Ethernet in the First Mile, it's an active market in France for example, we can get good connectivity, UK same sort of thing. But I've got an Ethernet node in Zambia; let me tell you there is no Ethernet in the First Mile capability in Zambia I have to buy fibre and all these sorts of things.
So it really depends where you are whether it got implemented or not and really what the local infrastructure was that was there to support it. But for us as a service provider, as a competitive service provider it's a huge enabler to bring large amounts of capacity to an end customer relatively inexpensively. Why did it take so long? We weren't there to push it maybe.
And maybe to add a bit to it, I think there's two dimensions I think one is Ethernet as a service as mentioned by James, and I think that's one dimension. The other dimension is Ethernet as a transport. And when you talk to service providers in those days we talked about okay we want to introduce Ethernet, and they said that's great we love it but we want restoration, protection and traffic engineering all the things that we are used to in the days of ATM frame relay, SDH Sonet, we want to basically recreate it on Ethernet. And we had to actually come up with new technologies, new standards and implement, get them deployed and that takes time. And I think that was one dimension of it. And at the same time there was the access part and getting basically services off the ground and the local loops etc, the service part is another dimension that service providers had to worry about.
Yali, I think you wanted to make --.
Yes, I just want to add to that. Definitely the infrastructure plays a role here, so in the case of Verizon in the US just this year we were able to launch 120,000 new links into buildings that's through GPON because we have the fibre network in there. So when you have the -- like the copper network then you don't the efficiency and the customer don't get all the benefits in the last number of years, so that's why probably the uptake was slower in the past but it will go faster and next year we plan to launch more.
I'd just like to speak up as a vendor of [inaudible] and the First Mile Solutions, we saw the Ethernet over copper as a piece of the overall puzzle rather than the end game. And what really bolstered the growth in Carrier Ethernet was the evolution of the work out of the Metro Ethernet Forum, or MEF and that's really because the services had to be defined and standardised. And we are really actually at a point now where things are taking off like crazy and it's partially because there is now a standard interconnect among operators.
And so at the external network to network interface it's been defined by the MEF, it's really enabling. And I think Yali you were talking over lunch that this is a huge enabler for Verizon as they move outside their traditional markets. So there is a big ecosystem component to this growth and what's taking -- what has taken so long I guess.
Definitely, just to add to that if I can that we -- in order for us to interconnect with the other carriers or use [us] as a Last Mile, then we have to certify each of those carriers. So the certification under the standards defined by MEF is very important for the carriers to work together and to go through the certifications.
Next question please. Oh we've got two more yes.
Hi. Okay, so what do you think is holding back Ethernet the most from all of you please?
Holding it back in comparison to what I suppose?
Well what I would say is that people are still choosing TDM technology, Sonet, DS1, even when Ethernet is available and that means something is holding back people from buying the Ethernet where they could. So what do you see is getting people to make the decisions to use like I say technologies even where Ethernet is available?
Human nature I don't want to change.
Yes, let me. So I asked our people what percentage of all the interfaces that we ship are Ethernet 90% plus, so I think the reality of your question is suspect. It's almost entirely Ethernet. Where there are -- where there is a legacy solution that runs -- somebody -- some carriers running over SDH or Sonet they may continue to do that, but the new shipments are not just a small percentage Ethernet, and it's Ethernet in different forms it's 10G, its 1G, its 100G but its Ethernet, it's that frame format. So I think your question is incorrect. It's overwhelmingly Ethernet.
I think just to add I think if there is one thing where we've been held back and we are now moving forward it's the wireless part, the mobile backhaul. I think when you look at standards like 2G and 3G they were not really optimized around packets and I think when [LTE] came around it is completely IP optimized for packets Ethernet and I think a lot of operators are migrating.
So if you look at for instance where a lot of these DS3s have been deployed it's been to base stations around the world to basically maintain the service of 2G and 3G. And that right now I think in the last few years has been accelerating. And if you look at mobile backhaul solutions and vendors it's a very interesting market right now.
Also in my view some countries the pricing makes a difference as well. For example in Canada the Ethernet price for 2M or 4M is higher than -- or 3M is higher than T1 or two T1s so the pricing plays a role here as well. And that's basically I see that in North America and amazingly not so much outside of North America.
Good answer. One more question here.
Who do you see as the [NYQ new entrants] or [inaudible] [SDN] do you see them like that, like a big evolution or a very important evolution for you?
Is SDN an important evolution for Carrier Ethernet?
Well I think that it's interesting because Ethernet is a protocol that's evolved without a control [plane]. So there are various ways to control and manage the Ethernet and native Ethernet services but usually it's inserted into something which in turn has a control plane like MPLS. SDN brings the possibility to have a native transport layer in Ethernet that is controlled by a higher order control plane, and I think that's an interesting innovation and it's something that even as a carrier we haven't fully thought through what the implications of that are.
But perhaps you could say that if an Ethernet frame originates from the back of the server and it can be transported all the way across the WAN as an Ethernet frame and you can have a single unified control plane that took it all the way from the back of that server all the way across the world to another location, that's a very significant evolution and we need to understand what the implications are of that.
I think something else we see within SDN from a carrier point of view is, and again this is something that we are still thinking about what the impact of it is going to be; SDN has got the potential to flatten the network economics model. Before we used to build networks for -- to meet particular price points, so we might build a network or an offering or something to produce something that's very, very low cost like for internet transport, for internet transit for example. So we carry 4,500 petabytes of traffic on internet every month over our network by using Juniper equipment.
We could not afford to do that if we had many, many, many layers of complexity that sat within in. On the other hand the services we provide on private MPLS or WAN Ethernet services have complexity in traffic engineering and latency concerns and class of service and prioritisation and latency -- strict latency controls, [fail over] and all that sort of thing that needs to be controlled and managed and that's very expensive.
If you are saying five years from now everything will have a common transport layer and there will be a common control plane and its routing layer then that says everything costs the same. And as a carrier we haven't thought through what the implications are of that. It costs exactly the same for me to carry an internet piece of traffic for somebody [per hours] on YouTube as it does to carry a trade against the New York Stock Exchange what implication does that have we still need to think.
Thank you. So I'm going to say thank you panellists. We've run out of time. And I would say press people go and nab them during the cocktail sessions okay, and ask them questions because they would love to talk to you. Thank you.