Q&A with Nortel’s PBT expert
If you follow Nortel a lot (or read Light Reading), you’ve probably hear us talk about a technology called PBT, or Provider Backbone Transport. PBT is a technology Nortel invented to make Ethernet more carrier-class for use in metro networks, and has sparked some industry debate because of it’s potential as an alternative to MPLS.
Nortel’s PBT expert - John Hawkins
Yesterday I talked to John Hawkins, Nortel’s resident PBT expert - who I’ve actually known personally for 7-8 years when we were both in Nortel’s Optical Ethernet team and working on various activities related to Resilient Packet Ring (RPR). Yesterday we talked about the difference between PBT and it’s cousin PBB, the industry debate between PBT and MPLS, Nortel’s Carrier Ethernet Ecosystem, and whether PBT was defeating the purpose of Ethernet. Below is our conversation.
What’s your title and role? I am senior marketing manager for Carrier Ethernet in the Metro Ethernet Networks Organization. My focus is evangelizing our Carrier Ethernet solution both externally to the industry and internally to our sales folks. I also support our marketing endeavors with IEEE, which is where PBT and PBB are being defined. And at the Metro Ethernet Forum I also do a lot of work with them to publicize the use of Ethernet in a carrier context.
What exactly is PBT, what is PBB, and what’s the difference? PBB stands for Provider Backbone Bridges. It’s what we used to call Mac-in-Mac. Technically it involves attaching an extra Ethernet header to a standard Ethernet packet. So you have an Ethernet header that belongs to the end user, and we put another header on top that one that becomes the purview of the carrier, or the backbone provider. That allows the provider to keep their customers separate and to differentiate types of traffic on the network. This in turn allows providers to scale networks in support of millions of customers or services.
PBT, or Provider Backbone Transport, is a way of interpreting the Ethernet header a little differently, but doesn’t modify the header at all. In a regular Ethernet switch, every time a packed comes in, the switch looks up the destination address to know where to send the packet. With PBT we pre-determine those paths and pre-provision them into the switch. So the switch still does the look-up, but the answer it gets on where to send the packet is pre-programmed. In this way we can create “deterministic paths” through the network - and they are well-behaved and predictable. And carriers like well-behaved and predictable networks.
But doesn’t that defeat the purpose of Ethernet? Not if you don’t throw the baby out with the bathwater. We can segment the switch so part of it can still behave like vanilla Ethernet. But parts of the switch that you need to be carrier-grade and have 50ms fail over and have these pre-determined, reliable, well-behaved paths can behave in the new way.
And where would they want that? They would want that anywhere they have a point to point trunk function - where they have already aggregated a lot of traffic and they have a high-bandwidth path. Wherever they are doing a trunking function and they really need a carrier-class connection. You can get these pre-determined, well-behaved paths today with other technologies like ATM and SONET/SDH but the advantage of doing this with Ethernet is that Ethernet switches are very cost-effective. And we’re not redefining the switch with PBT, so it’s pretty easy to teach a plain vanilla Ethernet switch to do PBT, and that gives you a great cost advantage.
