Cisco has a number of significant extensions to its venerable 7600 Ethernet edge router on tap next week at ITU. Among the more important is Broadband Remote Access Server (B-RAS), making the six-year-old workhorse Cisco’s strategic – for now – aggregation platform for video. Cisco Senior Vice President Mike Volpi shared his thoughts with Network World Managing Editor Jim Duffy on the ripened router, its ever-increasing roles and its competition.
Cisco has a number of significant extensions to its venerable 7600 Ethernet edge router on tap next week at ITU. Among the more important is Broadband Remote Access Server (B-RAS), making the six-year-old workhorse Cisco’s strategic – for now – aggregation platform for video. Cisco Senior Vice President Mike Volpi shared his thoughts with Network World Managing Editor Jim Duffy on the ripened router, its ever increasing roles and its competition.
Your share of the edge router market improved for the third quarter in a row, according to Dell’Oro Group. Are you doing something right, or are your competitors doing something wrong?
The edge market is a broad market because there’s everything from Layer 2 pure connectivity to Layer 3 richer services, everything from consumers to business users, and so forth. I think what we’ve executed well on the Cisco side is to cover all of these market segments, which has been helpful for us; and then we’ve had now two or three software releases that run on the 7600 which have really, really improved the carrier-class nature of the platform. Everything from manageability, particularly around Ethernet, support for carrier features like high availability, like rapid restoration, like VPLS/HVPLS, EoMPLS. But the net net of it is that we’ve got a lot of good offerings. And then at a high level, the kind of message we’re trying to convey to our carrier customers is that even though you’re migrating to Ethernet, the battle is fought on a unified edge, meaning you have one platform that can deliver multiple services, which means a services-rich edge platform which the 7600 is, as opposed to a lot of our competitors who tend to just think of Ethernet as a transport technology and don’t have a lot of rich Layer 3 functionality on the platform. I think that’s contributed to our success.
It seems the 7600 is focused specifically at Ethernet while you have other platforms – the 10000 series and the 12000 series – for B-RAS and multiservice functions. Is the 7600 really a converged platform then?
It increasingly is. We’re taking the B-RAS functionality and have ported it to the 7600. We’ll be introducing that in the first calendar quarter next year. We’re introducing it at ITU. We will have session border controller functionality on the 7600, we will have deep packet inspection functionality on the 7600, we’ll introduce richer QoS than the platform already has. So what we see essentially happening is that there is a market out there which is either pure B-RAS or pure multiservice edge, but what’s happening is the migration towards Ethernet. The way we look at it is, we have a platform that is very strong on Ethernet and we’re increasingly adding capabilities to it which are higher-level services. The 7600 roadmap is very rich in those sort of broad ranges of new services and a lot of our customers are increasingly buying that roadmap along with, of course, the existing Ethernet functionality on it.
Will those converged features be exclusive to the 7600 or do you plan the same for the 12000 and 10000 series routers?
No, the 10000 series is going to stay focused as a B-RAS platform; and the 12000 is going to be focused as the multiservice edge with orientation towards packet-over-SONET, and TDM hierarchy-based interfaces. From an Ethernet perspective, it really is going to be 7600 and our job is to add more and more functionality so that when customers buy the Ethernet edge, they also buy the richer functionality on it.
Analysts and other industry watchers continually point out the age of the 7600 as a drawback of the product. How were sales of the 7600 in Q3?
We don’t disclose sales per product line but it grew year-on-year in excess of close to 40%. So the growth rate is very good.
What analysts look at – they usually have two criticisms of the platform. One is that it’s “old”; what they actually miss is that the 7600 is a modular platform and since its introduction we’ve introduced three new supervisor (engines), and we’re on our fifth generation line card. It actually speaks to our ability to build a product that customers can install the chassis, and then as silicon technology and other technologies mature, we can incrementally enhance it over a period of time. So the 7600 today that we sell bears no resemblance to the one we sold six years ago. It’s still has the same model number but…very, very different platform.
The other thing that people say is well Cisco is developing other future generation products which, of course, we always are. We spend, just in service provider, in excess of $1.5 billion in R&D every year. So we are building next generation stuff but those things are two, three, four years out in some cases. The CRS-1 took us four years to develop. So we start it now and even after we introduce those products – like in the case of the CRS-1 and the GSR — we’ve added a tremendous amount of new functionality to the GSR after we’ve introduced the CRS-1. It’s not like you turn one on and turn the other one off. We’re always developing next-gen stuff. But that doesn’t mean the 7600 doesn’t have a rich roadmap that extends way beyond 2010 at this point.
So you’re now developing your next-gen Carrier Ethernet router?
Sure. Yeah, we are. I would characterize it as a different kind of a platform designed with more of a futuristic approach. It’s not really a successor (to the 7600) because successor implies that one replaces the other and they don’t exactly do that. We do believe that architecturally what our customers are going to do is this notion of a multiservice edge vs. B-RAS vs. Ethernet edge is going to eventually blur away towards a universal edge. And we’re developing such a platform. But at this point, we’re very focused on 76, and the 76 is going to have a very rich – in fact, an industry-leading – roadmap that goes beyond 2010.
What are your thoughts on Juniper’s MX960?
We always take it seriously. They’re a good company, they’ve generated good products in the past. It is a first attempt at a metro Ethernet product and at least as we can see it, it principally feels like a LAN switch in terms of queuing and buffering and sort of the general carrier-class attributes that are required for Ethernet. But it comes from a company that has some credibility in this business so we always pay attention to it. So far, where we see it competing is principally in a data center-type application rather than a true metro Ethernet application.
Do you see anything in the MX960 you need to respond to, or that exposes shortcomings in your products?
Not really. To be candid, we compete with other people more so in metro E. I think the MX will be a reasonable product. There’s nothing that sort of strikes us right now to say, ‘Boy, we just really missed that one.’ I think the principal strategy for the MX seems to be one of second vendor or third vendor (source). I think that’s the basic approach. I don’t know that there’s anything about it that truly stands out right now as being unique.
What about its form factor? Do you plan on coming out with an Ethernet router that approaches or exceeds 400Gbps in one-third of a rack?
I think if you actually look at our roadmap we have a fair bit of stuff that gets us in that direction. Although, candidly, if you heard my story about the Ethernet edge, density is one aspect of how a product might differentiate itself. We believe it’s more of those services at the Ethernet edge that will define who wins or loses. You can have the best density in the world but if you can’t do queuing for carrier-class features, and differentiate between voice, video or data, if you can’t have the appropriate control mechanisms like HVPLS or VPLS or MPLS, it doesn’t matter how dense your box is. You have to have that functionality to even offer the service. My view is, at least for now, the Juniper (MX960) product is missing a range of those features that are considered the norm in Carrier Ethernet.
Is the new B-RAS functionality on the 7600 in response to any specific RFP, like Verizon’s?
The Verizon RFP is certainly one that goes along the lines of this architecture. But this is designed because we believe that an approach of distributing B-RAS further to the network edge and integrating it into an Ethernet platform makes sense. The reason it makes sense is that, B-RAS is obviously used to launch not only classical DSL services but video as well. Video requires an Ethernet platform because of the amount of bandwidth it consumes. Right now, you’re essentially looking at a two-box solution with a fairly large number of standalone B-RASes not having the horsepower, the throughput capacity, to take the video bandwidth through it. By taking B-RAS and moving it to an Ethernet aggregation platform like the 7600, we essentially have the functionality of B-RAS running on a platform that has the throughput that’s capable of doing video. So Verizon is an example of it, but we think that any customer that considers really high bandwidth plus B-RAS, or video with a B-RAS control plane should think that this is the right technical approach.
Your B-RASes supports 64,000 sessions but Juniper’s been out there with the E320 for awhile at 128,000 sessions. Are you falling short?
First of all, we would like to see the 128K actually work. We’re not entirely convinced that it does. But that’s the difference between marketing claims and reality. Because we will have the 10000 (series) as a standalone B-RAS, the way we’re approaching the technical architecture of the network with the 7600 and the B-RAS is that by embedding B-RAS functionality in the 76, we’re aggregating video. Video has more bandwidth, therefore you can assume a lesser number of streams coming through the box. And you’ve got to keep in mind that these streams, if they’re video, tend to be constant bit rate which means you can’t take advantage of the statistical multiplexing. In all likelihood you’re not going to hold a 128,000 video sessions from a bandwidth point of view coming through a single product. So the reason for 64,000 sessions is that is the appropriate balance between session count and total throughput in the box when you’re trying to put video through it. So instead of maxing out on session count, which would not have done us any good if we put video through it, we’ve balanced between the bandwidth requirement of video and the session count that’s required on that type of an application.
We’re proposing a distributed B-RAS approach. By putting it on the 7600, you push it further to the edge so inherently you’re aggregating a lesser number of sessions into that particular B-RAS.
What is the future of ATM-based B-RASes with the advent of Ethernet B-RASes?
Virtually every single developer and manufacturer of DSLAMs at this point is utilizing Ethernet as their backhaul technology. So an ATM-based BRA-S is required to maintain what’s deployed today. So to a degree, you might need to have a little bit of ATM functionality on some of the next-gen B-RASes to aggregate the older DSLAMs. It’s purely maintenance of the installed base. We really don’t see a very big future for ATM-based B-RAS functionality.
What’s that mean for your 10000 series routers?
We’re putting a lot of Ethernet functionality (in it). We have an 8-port Ethernet card on the 10K… so we’re migrating it increasingly to an Ethernet-based approach. But obviously, we have that platform and we have customers that have it, and to the extent that they install Ethernet DSLAMs that aggregate on the 10K, but over the long term we would expect those customers to deploy 7600s with the B-RAS functionality on it.
Any particular timeline for adding more of the edge services into IOS XR you alluded to in March?
Essentially, we’re on six month software releases. We just introduced the XR 3.4 release about a month ago. That was fairly rich in terms of Layer 3 VPN functionality. Fairly higher layer edge services. Future releases will contain more layer 2 VPN functionality in them.They will contain session border controller functionality, firewalling and other encrypted tunnels like IPSec. So you can expect…every XR software release every six months or so we’ll be introducing a larger toolkit of edge features on IOS XR.
Eventually we’ll obviously have B-RAS on IOS XR but it’s not something that we’ve committed to a particular timeline yet. We have the code obviously, it’s a matter of moving it. We’ve moved it from one platform to another in the past. We know how to do it. It’s a matter of working priorities.
