Laurel Targets 10-Gig at the Edge

Laurel Networks Inc. is beefing up its edge router with high-speed interfaces. Today, the company introduced 10-Gbit/s Sonet/SDH (Packet over Sonet) and Gigabit Ethernet physical interface cards for the ST200 Service Edge Router.

The announcement represents an aggressive move by Laurel in the edge router market. Traditionally, service providers have used 10 Gbit/s in the core of their networks. Packet-over-Sonet OC192 connections and Gigabit Ethernet lines between core routers has become standard fare in many carrier networks. But until now, there has been little talk of edge routers touting 10-Gbit/s interfaces.

Analysts are divided over whether 10 Gbit/s is really needed at the service provider edge. Kevin Mitchell, an analyst with Infonetics Research Inc., says that denser access equipment and IP traffic growth is driving the need for 10 Gbit/s at the edge. He says that an edge device that can hand off 10-Gbit/s pipes to a core router is ideal, because it reduces or eliminates this aggregation from the core router.

On the other hand, Dave Passmore, research director for the Burton Group, disagrees: He says it will be at least a couple of years before carriers will need 10 Gbit/s.

“I’m kind of scratching my head on this one,” he says. “There isn’t much demand for 10 Gig at the edge right now. I see it more as a checklist item to show customers that the box can grow to more capacity.”

Laurel’s Sonet and Ethernet line cards support one port each, for a total of eight 10-Gbit/s connections per device. As part of the announcement, the company is introducing a next-generation version of its network processing blade. The older version handled only 5 Gbit/s per slot.

Steve Vogelsang, Laurel's vice president of marketing, says the new 10-Gbit/s Packet-over-Sonet interfaces will be used as uplinks to connect edge routers to core routers. The edge routers can then aggregate all the slower traffic and feed it into a 10-Gbit/s core, allowing carriers to eliminate a layer of aggregation.

The 10-Gbit/s Ethernet interface would be used to connect edge routers to core routers in the same point of presence. The reason to go with Ethernet in this situation is simple: it’s cheap. In general, Ethernet interfaces can cost 50 percent less than Sonet, says Vogelsang. Because the company still hasn’t published its pricing, he wouldn’t elaborate. Mitchell says that most point-of-presence connections already use Ethernet instead of Sonet as their transport technology.

The new line cards, which are expected to be available in the third quarter, will put Laurel’s ST200 in closer competition with high-end routers from Cisco Systems Inc. (Nasdaq: CSCO) and Juniper Networks Inc. (Nasdaq: JNPR), which both offer 10-Gbit/s interfaces on core routers. Cisco offers them on the GSR 12400, and Juniper has them available for the M160. Cisco also supports 10-Gbit/s interfaces on its 7600 edge routers. But Juniper’s edge routing platforms, the ERX and M-series, do not support either OC192 or 10-Gbit/s Ethernet. Laurel will likely still compete with Juniper, which has been pushing its M160 more toward the edge. Cisco has been doing something similar with the GSR 12400 family, as well (see Cisco GSR 12000 Sent to the Edge).

That’s not the only competition that Laurel may face. Cisco, Extreme Networks Inc. (Nasdaq: EXTR), Foundry Networks Inc. (Nasdaq: FDRY), Force10 Networks Inc., and Riverstone Networks Inc. (Nasdaq: RSTN) all make Layer 2/3 switches that support 10-Gbit/s Ethernet interfaces. Cisco and Foundry have recently announced new line cards for their 10 Gbit/s, while Riverstone has recently introduced a whole new platform (see Wavesmith: Giant Killer?, Foundry Intros Next-Gen 10-Gig Ethernet and Riverstone Fuels 10GigE Price War). Foundry and Extreme are both expected to announce new platforms this spring.

Vogelsang argues that Laurel is not really competing with these products, because they are strictly switching products. The ST200 is a full-blown router. Because these switches lack Packet-over-Sonet interfaces, they are not used in large carrier networks to aggregate access traffic into a 10-Gbit/s routed core.

The price points of the switches and the routers are also different. While Ethernet switch prices have fallen drastically over the last six months, 10-Gbit/s Ethernet routing interfaces have not benefited as much from falling component prices.

“You can slash prices on chipsets all you want, but our routing interfaces also include complex routing components that drive up the overall cost” says Vogelsang. “We certainly benefit from the higher volume production, but our interfaces will likely always be priced higher than a switching vendor’s prices.”

— Marguerite Reardon, Senior Editor, Light Reading

skeptic 12/5/2012 | 12:11:51 AM
re: Laurel Targets 10-Gig at the Edge Steve Vogelsang, Laurel's vice president of marketing, says the new 10-Gbit/s Packet-over-Sonet interfaces will be used as uplinks to connect edge routers to core routers. The edge routers can then aggregate all the slower traffic and feed it into a 10-Gbit/s core, allowing carriers to eliminate a layer of aggregation.
The reasoning here isn't very good. The
aggregation layer (if present) has to do with
the number of edge devices connecting to the
core device. And the number of edge devices
has to do with density of what the edge router
is connecting to on the access side.

The big uplinks only make sense if you are
consolidating to far fewer edge routers. Or
if there is a huge amount of traffic coming
into the edge. Neither makes much sense at
the present.

These interfaces provide extra flexability to
a small number of customers, but its not
huge need on the edge right now as far as
I can see.

leafy 12/5/2012 | 12:11:50 AM
re: Laurel Targets 10-Gig at the Edge Skeptic, have you ever done the math on this box?
They _are_ claiming to be able to reduce the number of edge boxes you deploy. You can easily fill this box up with more the 10Gbps worth of customer facing data.

Let say you put in 2 10Gbps cards and the rest are filled with customer facing cards:

(these number are taken from their web page, modified to account for the 2 slots being used for the 10Gbps interfaces)

24 GigE -> 24Gbps
224 OC3 -> 29.7Gbps
576 DS3 -> 25.9Gbps
4032 DS1 -> 6.0Gbps

So I guess you're correct ... if your going to aggregate _just_ DS1, but any realistic mix of cards will easily fill the 10Gbps pipes.
willrouteforfood 12/5/2012 | 12:11:48 AM
re: Laurel Targets 10-Gig at the Edge Leafy,
the reason it is overkill is the fact that SP's often design these aggregation links with roughly 5:1 oversubscription. I have seen more extreme cases as well. You have to take into account the laws of probability and the fact that you would never run the customer links so hot all at once. So, you are moving to a larger pipe that you will typically not be filling. This is not to mention the added OPEX costs associated with 10G links today (both POS and ETH). I agree that it is overkill.

Just one man's opinion,

dellman 12/5/2012 | 12:11:46 AM
re: Laurel Targets 10-Gig at the Edge Seems Laurel to wants to recapture headlines from procket.

Laurel supports more types of interfaces including ATM

From Laurel's website:

10,000,000 BGP RIB-In paths
1,000+ BGP peers
2,500+ OSPF instances
8,000 IP interfaces per line card
64,000 ATM interfaces per line card
64,000 VLAN interfaces per line card
64,000 Frame Relay interfaces per line card
1,000,000 global IPv4 routes per forwarding table
4,000,000 VPN IPv4 routes per system
16ms backbone fail-over for up to 512,000 routes

Could not find any scalability data about procket.

Cisco and Juniper should wake up and show how they can match..
Tony Li 12/5/2012 | 12:11:44 AM
re: Laurel Targets 10-Gig at the Edge I'll just point out that we've never seen a standardized, cost effective, high-speed link layer that did not get significant usage in the ISP market space. I also think that folks should look at the CapEx and OpEx costs of 10GigE very carefully before passing judgement. For intra-POP interconnect, I suspect that it will become the media of choice.

Of course, I'm just slightly biased. ;-)

Sign In