As you may have read elsewhere by now, Light Reading recently conducted the first independent test of the CRS-1 Carrier Routing System, the next-generation core router from Cisco Systems Inc. (Nasdaq: CSCO). The tests, conducted by European Advanced Networking Test Center AG (EANTC), followed a methodology crafted with input from a few select service providers.

Luckily for Cisco, the box passed with flying colors – no major claims were refuted, nothing caught fire, and no animals were harmed in the making of the test (see 40-Gig Router Test Results). Key elements run through the grinder included the device's 40-Gbit/s optical interfaces (the world's first) and the routing engine.

Light Reading ran this test because so much is at stake with the CRS-1. This is the product that will define Cisco's future with service providers, particularly as its IOS XR software replaces the old Internetwork Operating System (IOS) that runs on much of Cisco's carrier product line. Then there's also the hoopla that surrounded the CRS-1's debut in May, when Cisco touted the box as the heart of "The Next 20 Years" of the company (see Cisco Unveils the HFR).

Having peered into this supposed piece of history, Light Reading wanted to talk with the man in charge of the project: Mike Volpi, senior vice president and general manager of Cisco's Routing Technology Group. We'd gotten a chance to tap Mike's brain back in 2002 (see Mike Volpi). But this time we wanted to focus specifically on the CRS-1 – what it means and why Cisco is making such a big deal out of it.

Along the way, Volpi also talked about the potential revenues from the box and what Procket Networks Inc. means to the company (see Valley Wonk: The Procket Puzzle). Whether it all works out the way Cisco planned will take years to determine. For now, here's a peek at what Cisco is hoping for.

— Craig Matsumoto, Senior Editor, Light Reading

WEBINAR ALERTOn Thursday, December 9, at 12 noon New York / 9 a.m. California / 5 p.m. London time, during a free hour-long live Web seminar, representatives from Light Reading, EANTC, Agilent, and Cisco will present the results of the CRS-1 test and answer questions about them. Telecom Italia will also be participating.

To register for the December 9 Webinar click here.

Light Reading: We have to ask: Why did you allow Light Reading, of all publications, to do a test of this hugely important product?

Mike Volpi: Is this the part where I'm supposed to give a good plug for you guys?

Light Reading: Actually, the best plug for us is a really, really bad plug. Those are the ones people read and pass along to their friends.

Mike Volpi: We liked the fact that this involved actual testing as opposed to slideware and conversation. To show that what we talk about in our presentations and our slides is actually true, and have it verified by a credible third party – that was fundamental for us. Light Reading has a history of doing a lot of credible, valid tests, and that makes a difference in our view. And the reach of the publication within the industry was an important consideration, particularly amongst the technical community.

Light Reading: So what is your reaction to the results?

Mike Volpi: I've seen the summary of them. I guess my thoughts are that it performed more or less as expected, which is good. It performed at the advertised level with satisfactory results.

Light Reading: Can you pick any one aspect of the CRS-1 that you really wanted to see shine?

Mike Volpi: I don't know that there's specifically one. This is a product that had an ambitious range of goals. Certainly there were the performance goals, both in terms of straight throughput at the interface level as well as the aggregate performance in terms of number of interfaces and total packet throughput. We were pleased by those results. I was also pleased by the routing performance, where we were able to converge very rapidly on a large number of changes on a large route table.

The thing we've put a lot of design emphasis into, from a performance perspective, is the switch fabric – how it performs under load and how it deals with issues like multicast, which gets pretty complex in a distributed environment. That held up pretty well.

Light Reading: Now that the product has been in the public eye for a while, how do you distill it down? What are the aspects you point to first?

Mike Volpi: We have a number of customers that have tested it and a number that are in the process of deploying it in their active and live networks. The fact that we have a product that customers have paid money for at this point and are installing in their networks is obviously gratifying – the fruits of our labor, if you will.

Perhaps the most important thing is that with this product we were able to fortify the image of Cisco as a carrier-class company. Carrier-class means different things to different customers – scaleability; design and architecture of the product; the requirements for maintenance and upgrades; and so forth. When you talk to our customers [about the CRS-1] universally they come back and say, "You got it this time." I think we've erased any doubt in anybody's mind that we were committed to this space and had the skill set to compete in this space. At this point in time, that's the most important thing. A year from now we'll come back and say, show me the revenues, show me the earnings, show me the market share numbers, and so forth.

The other aspect to note is that when customers buy this product and evaluate it, they have a different impression of Cisco. They see a company that actually understands and works in the carrier segment effectively. It gives the customer a better feeling about our company, which allows us to win the marginal deals, the deals that are close, even if the product that's being sold is not a CRS-1.

Light Reading: You say you'll worry about it "later" – but how important is the CRS-1 in terms of revenue generation? Is it mainly a case of showing Cisco's potential, so you can sell services and other products into the network?

Mike Volpi: The core routing market as defined by Synergy Research or Dell'Oro Group is about $1.2 to $1.5 billion today, growing to maybe $2 or $3 billion in the next n number of years. If you look at that, relative to Cisco overall – we're a $22 billion company. Assuming we don't get to 100 percent market share, which is a fair assumption, you're talking at best $1 billion or $2 billion in revenue.

That's a wonderful number. We'd love to have an extra $1 billion in revenue through this product. But ultimately, the CRS-1's direct revenue impact at Cisco is going to be probably less than 5 percent. Maybe if we get really, really lucky, it's going to contribute 10 percent to Cisco's revenue. Is it a worthwhile investment in the context of the returns that it yields to us? Yes. Is it going to transform Cisco from a financial perspective? Probably not.

But I think it's a good investment, financially. And an even better investment if you think about the impact it has on the range of Cisco products.

Light Reading: Which is another topic we were hoping you'd discuss...

Mike Volpi: We built this product understanding that some of the components of it, whether they be software or ASICs or otherwise, would then be used in other products within Cisco. So, long-term, we've invested in something that yields benefits in several product lines, not just the CRS.

Light Reading: One component that interests us is obviously IOS XR, the carrier-class operating system. Is that something that is going to be retrofitted into other router lines?

Mike Volpi: It is definitely in the plan. We're not prepared to talk about which ones and when. We'll say more in the coming months.

Light Reading: How far is that going to go? Is this going to be the basis for what Cisco will look like, say, five years from now?

Mike Volpi: Not broadly. In the carrier space, it's going to be the core operating system for us. But you've got to understand, our company is very broad at this point. We've got WiFi access points and $300 routers, and it turns out IOS is a perfectly good operating system for those types of products. We're not planning to change any of those directions of the company. But at the carrier space, especially at the high end, we're going to be more broadly using this particular operating system software.

Light Reading: Wouldn't it fit at the high end of the enterprise, though?

Mike Volpi: To some extent. Is Microsoft going to deliver one or two very carrier-class routers in the middle of MSN? Yes, absolutely. But that doesn't represent the mainstream of the enterprise.

Light Reading: In talking about IOS XR, there is a lot of talk that what you've basically done is catch up to Juniper Networks and Avici Systems. How would you respond to that?

Mike Volpi: I would say that statement is coming from people that are not terribly informed about what we have. This is a truly distributed software architecture, which – I don't know what Avici has, but [IOS XR] is fundamentally different from Juniper's architecture. They run a BSD [Berkeley Software Distribution] UNIX kernel with a very limited number of modules that sit on top of it. Principally, the modules are large – all the routing protocols on a single module, for example – and the system is designed clearly in a master/slave or hierarchical architecture.

In contrast, ours is a truly distributed architecture, where you can have multiple router processors in the network all operating as peers – and where the modular granularity is very fine-grained, so BGP is one module, IGRP is another module, OSPF is another module, and so forth. Each can be upgraded or modified in a very fine-grained methodology. It allows us to create full logical separation, which you can't do with the Juniper architecture. All you can do there is virtualization of the routing table. So, it really is a huge step ahead in terms of the software architecture here. And for people that use it, that becomes especially apparent.

Light Reading: When do we get to see spinoffs of the CRS-1, like this half-sized CRS folks are waiting for? [See Sources: Cisco Building 'Son of HFR'.]

Mike Volpi: We're not going to announce new product lines [during this interview], but it's obvious to hypothesize that we're going to field different form factors and versions of this product. That's happened all the time at Cisco.

Light Reading: How important is the 40-Gbit/s capability as opposed to the 10 Gbit/s? Is that something people have really asked for?

Mike Volpi: It's important in the national research networks, and several of the fiber-rich large carriers, and we've had some good traction on that. I would say that in the very high end of the market segment, and in customers that tend to have a lot of capacity and are very fiber rich, 40-Gbit/s strikes a chord. Japan is a good example; where they have very hefty Tokyo-Osaka kinds of traffic, 40-Gbit/s is a capability they're excited about. It's particularly powerful when used in combination with something like the StrataLight Communications device, which takes 40-Gbit/s feeds and multiplexes them over a series of 10-Gbit/s channels. That's handy, because it allows carriers to deploy the 40-Gbit/s interface without having to change their optical infrastructure. We've had a lot of positive feedback on that.

Over the long term, it's going to get more important simply because of the amount of bandwidth that's continuously being required as a result of the broadband deployments around the world.

Light Reading: For the customers who are testing and deploying the CRS-1 – what are they looking for? Is it mostly the speed, the new OS?

Mike Volpi: It varies. A couple of things happen. One, RFPs come out. Those customers evaluate products on a broad range of criteria, particularly the forward-looking investment protection. In other words, when I buy this thing, how far does it scale, how long can they use it? They won't blast it with 40 Gbit/s today, but they ask, "Can I get to 40 Gbit/s when I need it?" That is definitely an important criterion – that's where the multi-chassis capability is important, too.

Other customers are just struggling with the amount of traffic they have to deal with. In those cases, rather than a full RFP process, they need a band-aid – something that's fast, that has the capacity.

Then there's another range of customers that just want to try out the operational environment, everything from the manageability to the operational knobs to the ability to configure logical or virtual routing. They're not actually going to put it in the network now, but they are planning their network design based on the CRS. And as a result of that, what they care most about are the software attributes of the product.

Light Reading: We used to hear there were very few people in the world who were true router experts – like, if you needed someone who really, really knew BGP, there were five of those guys in the world. Is that still the case? And in building the CRS-1, did you have to draw from outside Cisco to bring in some additional talent?

Mike Volpi: The pool is getting larger, although it's still in a relative sense very small. If there used to be five people, maybe that's quadrupled today. It's a select group. Many, many people understand how BGP essentially functions, but few of them know how it actually operates in a real-world environment. And in particular, what was tough on the CRS-1 was figuring out somebody that understood how it would work on the public Internet or in a large network and at the same time could design a distributed architecture, and implement BGP on top of that.

We did draw on a lot of talented folks in the company. Probably the one we profiled most was a guy named David Ward who is a Distinguished Engineer here at Cisco and who did a great deal of the architecture work for us around the protocols arena of XR. And there was an incredible team behind him. Some of them were just Cisco people that had been with us for a long period of time, some of them did actually come from the acquisition of IEng [Internet Engineering Group LLC, in 1999]. That was an important group of people that we brought in to help us with the BGP protocols. I want to emphasize I'm talking about the routing protocols here, not the whole operating system.

Most recently, we reinforced that group with a number of incredibly good people from Procket that we brought on board and added to that team [see Cisco to Pay $89M for Procket Assets]. The Procket routing team is now effectively part of the XR protocols and routing group, so we've actually fortified that group. We do believe routing knowledge is a limited skill set, and to the extent that we have our hands on those kinds of folks, that's a permanent competitive advantage.

Light Reading: What are the Procket people going to be working on?

Mike Volpi: The guy who ran the routing protocols group at Procket now runs the routing protocols group for XR, so we've injected a leadership person there. We do need a whole lot more edge features on the CRS-1, so they're running fast to deploy those features. And we've got more high-availability features and virtualization capabilities that we have on the roadmap – and, obviously, ports to other platforms. So, they're working on a broad range of topics there.

We have integrated them nicely now, to where about half the leadership of our XR software team comes from some of the folks at Procket, and the other half is Cisco. So at this point, there is no more Procket Group/Cisco Group. They're one organization.

Light Reading: I've also heard the Procket ASIC team is working on some jitter issues with the CRS-1, patching up some stuff.

Mike Volpi: Actually, the ASIC guys are not on the CRS-1. They're working on a different set of ASICs, future stuff that we can't talk about right now. They're working on a different program, actually.

Light Reading: Were any other acquisitions tied in with the 500 people who worked on CRS-1?

Mike Volpi: There are the people that designed the fabric, those are the guys from Growth Networks [see Cisco Buys Growth, Literally]. The overall router architecture is all Cisco folks. The overall IOS XR architecture is Cisco folks, with the exception of the protocols. That's pretty much it. So, Procket now, and the two past acquisitions, IEng and Growth.