& cplSiteName &

Juniper Divvies Up the Core

Craig Matsumoto
2/2/2009

Juniper Networks Inc. (NYSE: JNPR) is upping the size of its core routers, but more importantly, it's touting a carrier-network virtualization that could expand the market for monster multichassis systems. (See Juniper Updates Core.)

The idea would be to let carriers offer hosted backbone networks that all run on common routers. It's analogous to data center virtualization, where a particular server can host different applications for different clients. The idea isn't new to telecom, but Juniper thinks its technology, combined with carriers' need for new revenue sources, could really make it work.

"There are models that are going to take a while to get adopted, but the carriers we talk to, they see the potential," says Luc Ceuppens, senior director of Juniper's high-end business unit.

On the surface, the TX Matrix Plus -- which will be announced today and is due to ship in the third quarter -- is just another salvo in the war of the big core routers. Specifically, it's the box that makes multichassis routers out of Juniper's T1600s, a development that was promised when the T1600 launched in 2007. (See Juniper Attacks Cisco's CRS-1.)

So, let's start with the numbers.

The T1600 has 1.6 Tbit/s of capacity (800 Gbit/s if you don't add incoming and outgoing traffic together). The TX Matrix Plus can put 16 of the T1600s together to form a node with capacity of around 25 Tbit/s or, put another way, room for 1,024 10-Gbit/s ports.

The TX Matrix Plus still doesn't match the 92 Tbit/s claims for a maxed-out CRS-1 configuration from Cisco Systems Inc. (Nasdaq: CSCO). But it's a catchup step that Juniper needed to take.

"To be frank, they needed to bring the Matrix Plus to market, because it's something Cisco has had already," IDC analyst Eve Griliches says.

The argument for these multichassis routers has been that big carriers really are going to need that much routing firepower. Cisco says the multichassis market is starting to blossom, as it's got 20 customers using multichassis CRS-1s. (See Cisco Toasts Big Iron.)

That may be true, but it still defines a limited market for these boxes. That's where Juniper's virtualization pitch comes in.

The virtual world
Juniper's virtualization pitch plays on the idea that larger service providers already tend to use multiple subnetworks to accommodate services. "A lot of these networks are managed by different operational groups, and that's the way it used to be in the data center," Ceuppens says.

So, what could you do with carrier-network virtualization? Well, anyone remember Allegro Networks? (See Allegro Holds a House Party .) Similar to Allegro, Juniper is pitching its new core router as a vehicle for networking-as-a-service. A TX Matrix Plus setup would sit on the network as a pool of bandwidth that multiple parties could tap, letting a carrier lease out part of its backbone, in a sense.

For instance, large carriers could make peace with over-the-top video services by giving them their own content delivery networks, which the carrier could even manage. Verizon Communications Inc. (NYSE: VZ) has talked to content owners about this, Ceuppens says, and he thinks Google (Nasdaq: GOOG) is considering the possibility, too.

"A couple of years ago, everybody thought Google would build their own network. Now, I think they realize what an undertaking it would be, not only to build it, but to manage it," Ceuppens says.

Other possibilities: The carrier could become the wholesale provider of a small operator's network backbone, or lease out a network arm to a service provider that lacks a presence in a certain geographic area.

To Page 2

(11)  | 
Comment  | 
Print  | 
Newest First  |  Oldest First  |  Threaded View        ADD A COMMENT
Page 1 / 2   >   >>
OldPOTS
OldPOTS
12/5/2012 | 4:12:46 PM
re: Juniper Divvies Up the Core
Ah, the risks of packet switching networks carrying competitors. The choices of routing.

OP
Stevery
Stevery
12/5/2012 | 4:12:42 PM
re: Juniper Divvies Up the Core
I think I'm not getting the point of network virtualization.

For instance, large carriers could make peace with over-the-top video services by giving them their own content delivery networks, which the carrier could even manage.

So pretend I run a youtube clone. My choices are to

a. Pay for bandwidth at various data center entry points.
b. Pay for a virtual network service.

What _exactly_ makes me choose b over a?
Light-bulb
Light-bulb
12/5/2012 | 4:12:39 PM
re: Juniper Divvies Up the Core
I'm no expert on a CRS-1, but all the information I have already points to having this ability innately on the platform. Control plane is said to be able to be shared across linecard chassis, and there is no separate device (JCS) that has to manage this. So isn't SDR a hardware virtuilzation? As it requires separate linecards and RPs? So all control plane is completely separate?

I'm curious, if Juniper requires a JCS to manage all the "secure" virtualization entities... what happens if the JCS needs to be power cycled, or if it's inadvertently taken offline. (Never happens) Would the whole system go down?

Good to hear that they are validating another vendor's strategy though, perhaps there is hope to have some real competition out there.

Though I was shocked... (SHOCKED) that Infinera thought IPoDWDM just isn't a good idea... jeez no conflict of interest? Personally I think the marriage between ADM's and Routers needs to happen sooner rather than later, and intelligent control planes to help facilitate the change of color on the Router optic or a new provision of service across a multi-degree ROADM network has some very cool benefits.

Cheers
cfl-bulb
cfl-bulb
12/5/2012 | 4:12:38 PM
re: Juniper Divvies Up the Core
I have been hearing about 92Tbps multichassis from Cisco for quite sometime. What I know is that they currently deliver a 4 CRS-1 multi-chassis system that delivers around 5Tbps. Am I missing something or the analyst and others cannot figure facts from slideware.
Light-bulb
Light-bulb
12/5/2012 | 4:12:37 PM
re: Juniper Divvies Up the Core
Would 5Tbps make it still the largest currently available router in the world? Or is there another vendor who can deliver bigger? But yea your right cfl, Cisco has been saying that it would scale to 92Tbps, but then again what something is designed to scale to doesn't mean it's currently shipping. I wonder if it's more of a request thing from customer, and currently no router needs to be that large.
Personally I'm looking forward to hearing about the TX, as I'm not sure how it will work with priority multicast, and the depths of the Fabric. That's one thing I'd sure like to see tested is how these big behemoths work with actual loads etc. Personally I'm a fan of implementing QoS throughout the network, from CE to CE. Though some prefer a Crunchy on the outside chewy in the middle concept. (QoS at the edges, stupid pipes on the inside)

As anyone tested these large systems with real-world use? I.E. Cable, Tripe/Quad-Play, etc? How do these things work with large multicast flows etc.

Hey cfl-bulb... I hope you are mercury free... and real nifty name. :)

Cheers,
cfl-bulb
cfl-bulb
12/5/2012 | 4:12:36 PM
re: Juniper Divvies Up the Core
Would 5Tbps make it still the largest currently available router in the world? Or is there another vendor who can deliver bigger?

I was trying to understand if there is any difference between Cisco 92Tbps and Juniper 25Tbps announcements. Are they both statement of possibilities or is there a difference?

Hey cfl-bulb... I hope you are mercury free... and real nifty name. :)

Yeah, you need to be in tune with the economic and environmental needs.


OldPOTS
OldPOTS
12/5/2012 | 4:12:35 PM
re: Juniper Divvies Up the Core
"Personally I think the marriage between ADM's and Routers needs to happen sooner rather than later, and intelligent control planes to help facilitate .... a new provision of service across a multi-degree ROADM network has some very cool benefits."

That was my thoughts as I did my post, but failed to say it cutely.

LB How long will your filament last?

OP
OldPOTS
OldPOTS
12/5/2012 | 4:12:34 PM
re: Juniper Divvies Up the Core
If you properly divvy out functions between (JCS) controller and line cards the line cards can continue to run based on last directions until control is established (after power cycle).

Then you can run parallel controllers, even in different chassis!

Done by another vendor many years ago. They just don't have the speed yet.

OP

OP
Pete Baldwin
Pete Baldwin
12/5/2012 | 4:12:32 PM
re: Juniper Divvies Up the Core
The 92 Tbit/s is Cisco's theoretical maximum for the CRS-1. It takes 72 chassis to do it. It wasn't possible to do on the day of launch (in 2004), and my impression is that it's still not feasible now. I'll have to do some homework on that. In any event, it's a long-long-term number.

It's not clear to me whether Juniper can do 25 Tbit/s from the get-go, but i'd guess they can. It only requires 16 of the T1600s to be put together -- much more reasonable in terms of logistics alone.

I wonder if, by the time that figures like "92 Tbit/s" become useful, carriers will have decentralized the core. That seems like an awful lot of traffic to put through one spot in the network (even if we're talking about a virtual "spot" that's split among COs or even cities).
Pete Baldwin
Pete Baldwin
12/5/2012 | 4:12:31 PM
re: Juniper Divvies Up the Core
Thanks to Om Malik for giving us a friendly shout-out on GigaOm:
http://gigaom.com/2009/02/02/w...

... which got carried over into the NY Times:
http://www.nytimes.com/externa...

(Exact same content on both links, except the surrounding ads, of course.)
Page 1 / 2   >   >>
Featured Video
Upcoming Live Events
November 5, 2019, London, England
November 7, 2019, London, UK
November 14, 2019, Maritim Hotel, Berlin
December 3-5, 2019, Vienna, Austria
December 3, 2019, New York, New York
March 16-18, 2020, Embassy Suites, Denver, Colorado
May 18-20, 2020, Irving Convention Center, Dallas, TX
All Upcoming Live Events