Core routers are smarter, more modular, reliable

Beyond the enormous advancement in forwarding speed, today's core routers are more reliable, more intelligent, and feature increased modularity over their previous incarnations—good news for carriers hoping to capitalise on the economic potential of IP/MPLS networks. Today, about 80% of carrier revenue comes from voice rather than data services. Carriers would like to drive that ratio closer to 50-50 using a converged network that would support ATM, frame relay, even circuit-switched services.

For carriers, "it all comes down to one thing," says Carey Parker, vice president of marketing at Chiaro Networks (Richardson, TX). "Do you have or can you make a reliable IP/MPLS core router? If you can, then this convergence can happen."

Previous-generation core routers provided three-nines reliability or an unacceptable 10 hours of downtime per year. Today's routers must approach the degree of availability associated with traditional TDM networks, which is five-nines availability. To that end, the Internet Engineering Task Force is working on a suite of protocol-specific standards, generally known as "Graceful Restart," that will enable the seamless protection of routing networks. "Whenever you have failures within a router, it should be transparent to the rest of the network," explains Parker. "The reason this has been so difficult in the core-routing space is there are a lot of technical issues, primarily involving the software piece. You just can't quite get it where it's always transparent fail-over."

Of course, reliability is just one hurdle facing core-router vendors; they also have to consider the economics of the device. "Service providers are really pressing vendors to make the technology relevant in one of two ways: Either use the technology to increase top line revenue and profitable services or to reduce the total cost of ownership," observes John Doyle, director of product and technology marketing for Cisco Systems' (San Jose, CA) routing portfolio. "If the technology doesn't pass one of those two litmus tests, it really isn't of tremendous interest to the service provider."

Most of the RFIs and RFPs put out today focus on longevity, as carriers no longer have the luxury of an 18-month lifecycle. Now, they want to see 7–10-year lifecycles and 99% capital preservation. "You can't just churn through the network," says Parker. Instead, routing equipment "must grow like a platform. What that means is that you have to have a true multichassis design."

Historically, multichassis configurations have been too difficult to configure; they didn't scale very well and were based on proprietary technologies. However, carriers today are demanding this kind of configuration to increase their forwarding capacity without increasing system complexity. A multichassis configuration enables the service provider to have a single point of control for multiple chassis of line cards, which helps reduce operational overhead. Carriers also like multichassis or modular systems because they can reuse individual routers for various applications. "A router has to be flexible enough so you can use the device in a number of different roles in the network," says Matt Kolon, senior solutions specialist at Juniper Networks (Sunnyvale, CA). The new-generation routers might be used as aggregation routers or even edge routers if they feature the required functionality.

To make core routers even more easily upgradeable, several vendors are now "more embracing of network processors"—a move made possible because "network processor vendors are closing the performance gap on custom-made ASICS," reports Mark Bieberich, senior analyst at the Yankee Group (Boston, MA). "Network processors drive development costs down and add more flexibility to the product. I think it's an important trend to keep an eye on."

In the past, it was generally assumed that core routers were "dumb, fast things in the middle of the network and all the intelligence of the network was accomplished by the boxes at the edge," notes Kolon. The latest-generation core routers are disproving this old assumption—and will continue to do so.

"The success of the Internet is predicated on the 'inter' part of things," Kolon muses. "Where carriers are looking to take this next is to be able to offer [new] 'inter' things. We have the Internet, but how about 'inter' VPN [virtual private network], the 'inter' telephony network, or the 'inter' videoconference network? Carriers want to address these business opportunities that come from inter-carrier services."

To do this, future core routers will need to separate traffic into the various services and treat each accordingly. They must be able to communicate over service-provider boundaries discreet information about the traffic passing back and forth. They will also need to be able to find and report on the different types of traffic for billing purposes. Increased functionality and greater intelligence within core routers is "the only thing that is going to allow us to move the service-provider industry and what people think of as the Internet to the next phase," advises Kolon.

The carrier community certainly seems to be moving forward. There are three RFI/RFP/RFQ lab activities ongoing in Europe, four in North America, and several in Asia, reports Parker, who adds that all three continents are "amazingly in sync on this particular issue."

Only a few core-router vendors will benefit from this activity, however. After 2004, says Bieberich, the market will not be able to support any more than four vendors. Cisco and Juniper are the incumbents, and together they control most of the market. Vying for the remaining market share are Avici Systems (North Billerica, MA) and startups Chiaro Networks, Procket Networks (Milpitas, CA), Caspian Networks (San Jose), and Hyperchip (Montreal, Canada).

Avici just recorded a strong quarter, says Bieberich, but it "can't survive with just a 2.5% market share at the rate [it is] burning cash." For the company to survive, it will have to make the most of its recently announced partnership with Huawei. Similarly, Procket and Chiaro—while "the strongest of the four startups"—will also need to position themselves "either for multiple strategic alliances or for acquisitions," he reasons.

More in Network Design