A new MPLS debate heats up

Increasing interest in provider backbone transport, or PBT, technology as a simpler, cheaper alternative to multiprotocol label switching, or MPLS, has given rise to a new PBT-based optical transport platform, igniting new debates over metro and core technologies.

After declaring PBT its metro technology of choice last summer, Nortel Networks claimed vindication last month when British Telecom selected PBT gear from Nortel and Siemens for its next-generation network. Nortel said then that PBT in the metro and edge was a good complement to MPLS in the backbone. But it also admitted it was already talking to some carriers about deploying PBT in backbones as well.

This week Meriton Networks threw its hat in that game, promising to add PBT to its 7200 optical switching platform in the second half of the year. While Nortel built PBT into its metro Ethernet switches, Meriton combined PBT with DWDM to create an optical switching platform it calls “carrier Ethernet transport.”

As carriers move to replace inefficient Sonet with packet-based networks, they still want some of Sonet's connection-oriented characteristics, like the security of protected paths through the network. Based on the IEEE's 802.1ah standard, PBT sets up point-to-point Ethernet tunnels by affixing media access control (MAC) headers atop packet-based traffic. Rather than perform deep packet inspection, metro switches using PBT need only keep track of their own MAC addresses, making it a relatively simple method.

Meriton is contrasting its approach with the IP-over-DWDM (IPoDWDM) architecture Cisco Systems added to its CRS-1 core router more than a year ago. Unlike IPoDWDM, the beauty of Meriton's method, according to the company, is that it keeps the service and transport layers separate, minimizing the number of handoffs between the two and thereby simplifying the network. The optical transport layer is kept relatively unintelligent, burdened only with enough Layer 2 functionality to establish sub-wavelength traffic tunnels.

“Think of it as Layer 2 light,” said Bill Gartner, Meriton Networks' chief operating officer. “Much of the Layer 2 processes, like spanning tree, are turned off.”

When optical transport is based instead on a direct connection to intelligent routers, as in Cisco's IPoDWDM approach, it adds complexity and cost to the system. “It means you buy a lot of routers,” said Michael Howard, Infonetics Research principal analyst.

At the same time, PBT currently has limited scale in the metro, some analysts say. Though it theoretically allows for some 16 million connections, they have to be assigned manually, which can be tedious. “Carriers need to essentially know where they want the connection,” Howard said. “They have to specify it. There's no automatic traffic flow like a router could figure out. In the future, I'm sure someone will figure out how to automate it.”

Carriers that have already invested in a lot of routers and MPLS may be more open to IPoDWDM. But if they don't embrace it broadly, Howard expects Cisco, in the face of PBT's rising appeal, to at least introduce a next-generation platform that combines Ethernet and WDM more efficiently than its current Sonet-to-Ethernet migration box — the 7-year-old 15454. And Nortel, whose current PBT offering is based on metro switches and edge routers, is likely to introduce its own PBT optical transport platform to match Meriton's.

Carriers that embrace PBT in their edge or metro networks probably won't stop there, deploying it in their backbones as well, Howard said. “The steady march of Ethernet is not going to stop in the metro. It's going to go on to long-haul transport.”