Ethernet on a Grander Scale

The takeaway from Supercomm 2004 boils down to these three phrases: all things Ethernet; triple play; and cautious optimism.

Yes, there were other topics as well, but Ethernet was certainly one of the topics on everyone's lips. It's incredible that this technology has made the transition from the enterprise LAN to a ubiquitous user interface to the WAN – and this process is entirely driven by customer demand. Of course, Ethernet services have been around for many years; BellSouth Corp. (NYSE: BLS), for one, will soon be celebrating the tenth anniversary of its launch of Ethernet services.

But all those first-generation Ethernet services operated over a Sonet/SDH infrastructure. Each Ethernet connection would initially be given some chunk of bandwidth carved out of the TDM core. The services were point-to-point – essentially a cheaper replacement for a leased line, allowing the enterprise customer to use a cheap Ethernet port on a router instead of an expensive WAN port.

Ethernet services have come a long way since then. Sonet/SDH multiservice provisioning platforms (MSPPs) now typically include an Ethernet switch, or even a full IP router to front-end the time slots. But since Ethernet is a packet service, it makes sense, in the long run, to offer it over a packet infrastructure.

Heavy Reading's latest report – "Ethernet Over IP/MPLS Service Delivery Platforms: A Heavy Reading Competitive Analysis" – focuses on technologies that deliver Ethernet service over packet-based (IP and MPLS) networks. While Ethernet itself offers a simple and high-capacity interface for bandwidth, service providers ultimately have to make a decision about how they will provision and scale their services.

There are a couple of high-level trends to which we can point. The first is that MPLS is regarded by most vendors as the optimal multivendor service-provisioning technology. The second is that it's the right technology to help Ethernet scale beyond the metro network. For almost all of the vendors we surveyed, the specific solution is to use Virtual Private LAN Service (VPLS).

In terms of VPLS leadership, Alcatel SA (NYSE: ALA; Paris: CGEP:PA) and Riverstone Networks Inc. (OTC: RSTN.PK) have driven development of the standard from the beginning. Cisco Systems Inc. (Nasdaq: CSCO) is now coming up fast, after initially approaching the technology skeptically. My guess is that a lot of Cisco metro Ethernet customers will end up running this service over IP tunnels or tag switching for a while yet.

Extreme Networks Inc. (Nasdaq: EXTR) and Foundry Networks Inc. (Nasdaq: FDRY) are coming out of the enterprise and data center worlds with switches adapted for carrier operations. Both companies are proposing Layer 2 architectures in the metro, while turning to MPLS to scale service out to the WAN.

Speaking of the WAN, we're seeing specific Ethernet service features being introduced into products from Juniper Networks Inc. (Nasdaq: JNPR) and Laurel Networks Inc. Both of these companies are working their way out from the core/edge and now have smaller products that can be deployed as metro service-aggregation platforms, but both companies see this part of the network as belonging to the Layer 3/MPLS world.

There is also a lot of crossover of Ethernet services from the optical world. Atrica Inc. has a great solution for Ethernet over DWDM, and Corrigent Systems Inc. is offering an Ethernet-over-RPR solution.

Overture Networks Inc. offers a very impressive range of entry-level boxes, targeted at North American enterprises, the majority of which don't have access to fiber, leaving them to get their next-generation services over a T1 or T2 line. Surprisingly, Tellabs Inc. (Nasdaq: TLAB; Frankfurt: BTLA) has fielded a solution that looks good, and it will be interesting to see what kind of market traction it can achieve.

All of these companies are committed to MPLS as the way to support Ethernet in the WAN.

The big exception in the survey is Nortel Networks Ltd. (NYSE/Toronto: NT). It is pushing a Layer 2 end-to-end solution that uses a proprietary encapsulation scheme at the Ethernet level to solve the two problems mentioned earlier – service provisioning and scaleability. Nortel doesn't move to MPLS until beyond the core/edge.

Exceptions aside, however, we can say that there is now a distinct market segment of switches that offer Ethernet services over packet infrastructures. And there's a more focused consensus emerging about the properties of those infrastructures – they will tend to be Layer 2 at the customer, but the MPLS demarcation point is moving closer to the customer all the time. They will be multipoint-to-multipoint, as well as point-to-point, networks. VPLS is the primary technology to make the Layer 2-to-MPLS connection, so MPLS becomes both the service-provisioning technology and the scaling technology, allowing Ethernet services to achieve a global reach.

— Geoff Bennett, Chief Technologist, Heavy Reading

Sign In