Juniper Attacks Cisco's CRS-1

With the T1600, Juniper hopes to outdo Cisco's core-router density and add more service awareness to boot

Craig Matsumoto, Editor-in-Chief, Light Reading

June 11, 2007

5 Min Read
Light Reading logo in a gray background | Light Reading

Juniper Networks Inc. (NYSE: JNPR) is launching a direct strike at the Cisco Systems Inc. (Nasdaq: CSCO) CRS-1 Carrier Routing System with a new core router aimed at giving service providers more control over the traffic that traverses the IP network.

The new T1600 claims to pack a higher-density wallop than the CRS-1. But it might be the services angle that gives Juniper an edge, opening the old debate over how much intelligence should be placed in the core of the network.

"There's a trend here toward more service awareness and policy in the core. It's kind of a departure from history," says Shailesh Shukla, Juniper's vice president of service provider marketing and partnerships.

Juniper's been suspected of working on a new core router. (See Report: Juniper Rethinks the Core.) But analysts didn't expect it to arrive this soon, considering Juniper typically won't announce a product until a customer can be announced, too.

The company appears to be breaking that rule this time, although Matt Bross, CTO of BT Group plc (NYSE: BT; London: BTA), is expected to give the T1600 some moral support in a press release today.

The T1600 won't ship in production until the fourth quarter, Juniper says.

By complete and total coincidence, we're sure, Cisco is announcing today that it's shipped 900 CRS-1s to more than 85 customers including BT. Cisco is also revealing that Neuf Cegetel Group (Euronext: NEUF) is its latest CRS-1 buyer. So it claimed the early lead and doesn't appear to be slowing down. (See Cisco Pushes CRS-1 and Neuf Cegetel Uses CRS-1.)

Dense and Denser
When it was released in 2004, the CRS-1 was touted as the kind of router that could sit in a provider's network for decades. Part of the idea was to make the thing scalable -- allowing multiple chassis to be strung together as one router with a capacity of up to 46 Tbit/s. (That's 92 Tbit/s if you count both incoming and outgoing traffic, as router types often do.)

When it comes to one seven-foot rack, Juniper appears to have outdone Cisco. The T1600 fits in half a rack and provides 800 Gbit/s of capacity (not using the industry's double-counting convention). The largest model of CRS-1 fills a seven-foot rack and provides 640 Gbit/s of capacity.

The T1600 can't yet outdo 46 Tbit/s, though, so Cisco's world record is safe. (See Cisco Grabs a Guinness.) A Juniper spokesman says the T1600 will support multichassis configurations later; until then, anyone needing that density will be pointed to Juniper's T640 and TX Matrix combination.

One neat trick about the T1600 is that it uses the same chassis as the T640. This means it's possible to upgrade a T640 to the T1600 by just swapping out the switching cards and power modules, a process that Juniper claims takes just 90 minutes. (You do have to upgrade to the latest version of JunOS software, too.)

That's going to give Juniper an opportunity to poke at the CRS-1, which uses hardware and software different from Cisco's 12000 series routers.

Hardware advantages come and go, though. Juniper's bolder step with the T1600 lies in the amount of service awareness baked into the box.

The Intelligent Core
Arguing that carriers want more control over IP traffic, Juniper built the T1600 to work with the company's session and resources control (SRC) software. In the case of an all-Juniper network, the SRC becomes an eye in the sky, coordinating a service's path across the entire network. (See Juniper Pushes Services Policy.)

This means Juniper could set up a path for a particular video or VOIP connection, with the SRC telling routers to adjust bandwidth and QOS properly to make sure every service gets to its destination unhindered.

Any router implementing MPLS can do something like this, but Juniper says MPLS alone wouldn't be as effective. "You get taken by surprise a little more readily than if you have the policy engine taking control of the network," Shukla says.

This attention to policy would seem to contradict the usual model of a dumb core network, one that's made to shovel masses of packets quickly while the edge routers do all the thinking.

"Over the past four years, people have been moving intelligence out of the core," Basil Alwan, president of the Alcatel-Lucent (NYSE: ALU) IP division, told Light Reading recently. But Alwan, speaking before the T1600 became public, was talking about relegating full IP routing to the edges and using MPLS in the core; this arrangement doesn't necessarily conflict with the way Juniper is using the T1600.

Cisco has been pitching service awareness for its routers, but primarily at the edge, in the 7600 series. For the core, Cisco boosts QOS by using secure domain routers, where important traffic (video, for instance) gets segregated within a core router. This keeps other services from hogging the bandwidth required by the video streams.

Tom Nolle, president of consulting firm CIMI Corp. , thinks the idea of a policy-sensitive core could catch on. Carriers don't like the idea that their "premium traffic is just floating around out there," he says. "What you're going to see in the industry is an increased drive by the IP players to deal with premium traffic without having everything fall into the best-effort hole."

If equipment vendors don't make that move, Nolle thinks they'll risk having carriers "flee IP" in favor of Layer 2 options like Ethernet with Provider Backbone Transport (PBT). One reason PBT gets attention it does is "because the IP community has not wanted to get involved in end-to-end handling."

Some more specifics about the T1600: Each line card includes two Juniper-designed packet processors that can forward 50 Gbit/s apiece. Juniper claims the box is ready to handle 100 Gbit/s per slot when that time comes.

But the T1600 also accepts any of the T640 cards. So, in the aforementioned upgrade, a customer could upgrade to a T1600 while leaving all of today's line cards in place.

— Craig Matsumoto, West Coast Editor, Light Reading

Read more about:

Omdia

About the Author

Craig Matsumoto

Editor-in-Chief, Light Reading

Yes, THAT Craig Matsumoto – who used to be at Light Reading from 2002 until 2013 and then went away and did other stuff and now HE'S BACK! As Editor-in-Chief. Go Craig!!

Subscribe and receive the latest news from the industry.
Join 62,000+ members. Yes it's completely free.

You May Also Like