Nortel will enhance a proprietary technology for improving the performance of Ethernet as part of its emphasis on becoming a leading supplier of metro Ethernet systems for carriers.
Nortel will be adding link state protocols and synchronization to Provider Backbone Transport (PBT), a Nortel-developed derivative of the Ethernet standard intended to bring connection-oriented characteristics and deterministic behavior to Ethernet. PBT turns off Ethernet's Spanning Tree and MAC address flooding and learning characteristics to enable Ethernet to behave more like a traditional carrier transport technology.
Nortel recently announced a strategic plan to focus on metro Ethernet as one of the key market drivers for the company's future growth. Market tracker Infonetics Research predicts the Ethernet services market will reach US$22 billion in 2009 from US$2.5 billion in 2004.
The addition of link state protocols to PBT will enable carriers to configure Layer 3 topologies within their metro Ethernet/PBT network, said John Roese, Nortel CTO during an interview last week at the company's Ottawa labs. A Layer 3 topology will make the Ethernet network look and behave more like an OSPF or IS-IS router network, in which each router broadcasts or multicasts information on the cost of reaching each of its neighbor nodes.
This is what carriers prefer, according to Roese, adding that it will also facilitate multipath trunking. Multipath trunking allows operators to configure multiple links between switches and servers instead of traditional point-to-point trunks, which in turn enables network configurations for optimal traffic handling and load balancing.
"We've realized that the topology requirements of the carrier are much more sophisticated: they like the idea of Layer 3 topologies because it gives you the ability to have multipath, and more intelligent multipath," Roese said. "While you could try to use things like (IEEE) 802.1s and multiple Spanning Trees, that's not really desired by the carrier to be running Spanning Tree across their network."
Link-state PBT is in Nortel's labs. It is unclear when it will emerge in PBT-enabled products, which currently includes Nortel's Metro Ethernet Routing Switch 8600 and next year the Optical Multiservice Edge 6500 provisioning platform.
Adding link state is nothing new to Roese. While at enterprise LAN vendor Cabletron Systems during the early 1990s, Roese developed SecureFast, the company's virtual LAN technology. SecureFast was essentially Layer 2 OSPF, he said.
Synchronization will bring just that to an Ethernet/PBT network -- the ability for data transfers and multiple networks to operate in synch according to a clocking mechanism, thereby reducing the amount of buffering required between each element in the network. Ethernet is currently asynchronous but timing will make it behave more like SONET transport networks, which are ubiquitous in telecommunications.
"You can...actually have a Stratum 3 clock sitting in the core of your network somewhere and have consistent clock across that network that can be extracted by the endpoints," Roese said. "Ethernet is getting very close to what the SONET world could actually deliver in terms of capability but it's doing so at a price point that's orders of magnitude lower and a simplicity that is tremendous."
Nortel has synchronized Ethernet/PBT running in its Ottawa labs but product plans are as yet unclear.
Nortel is also positioning PBT as an alternative to MPLS and Virtual Private LAN Services in the metro Ethernet network. MPLS was initially developed 10 years ago to accelerate core routing but began to take on a Jack-of-all-Trades role, repeatedly pitched as a solution for traffic management, QoS, virtualization and other challenges, Roese said.
Along the way, MPLS was continually being displaced by other advancements better suited for those tasks, he said.
"I hate to describe MPLS and VPLS as a solution in search of a problem, but it's a classic example of how a technology that was designed for one thing iterated so many times that it lost sight of what its original purpose was, which was actually to improve the performance and capability of the network," Roese said. "There's was almost an irrational behavior to try and push those technologies to places where they had no business. This thing has taken a life of its own, it's been aligned to problems it was never meant to solve, and now what we're trying to do as an industry is unwind some of that."
Carriers BT and Shanghai Telecom plan to deploy PBT in their next-generation networks. And Roese says PBT is even winning over some converts among Nortel competitors.
"The people who are the biggest anti-PBT people on the planet -- our friends at Cisco and other folks -- are now coming out and saying, 'Well, you know what: We're going to do this too. And we're going to use the standard,'" Roese said. "Well, the problem with their dialogue is that they're kind of a little late to the party; I'm not sure they're sincere about it, to be perfectly honest; and the reality is, what we've learned is that the standardization of the elements of PBT that are in IEEE 802.1 are not the complete picture."
Neither is PBT, according to Cisco.
"PBT is not a standard and is just one of the many approaches being suggested for Ethernet transport as an alternative to Queue-in-Queue, MPLS Pseudowires, VPLS, etc.," a Cisco spokesman says. "PBT is statically configured and relies on an external (network management system) and provisioning tools to calculate and re-route around failures. While PBT utilizes 802.1ah for it's forwarding plane - they are not equivalent. Cisco supports 802.1ah and the scaling it provides for Ethernet based services.
"Cisco...supports Ethernet (operations, administration and maintenance) today across our routing portfolio," the spokesman continued. "Together these functions form the Ethernet foundation layer similar to PBT. We'll continue to work with customers on the evolution of Carrier Ethernet in relevant standards bodies."