gdt | Provider Backbone Transport (Reply)

MPLS is a nice technology. It allows the network topology to be altered using tunnels, but without the drama of IP in IP tunnels. MPLS can be used to solve some otherwise tricky network engineering problems, such as excluding some classes of customer from some links.

But then people had to apply MPLS to virtual private networks. VPLS is a design disaster, routers simply shouldn't be holding MAC tables and the entire technology has the feel of using a sledgehammer to crack a nut.

Provider Backbone Transport looks like a much better approach to implementing VPNs. Basically a VPN is a VLAN for that customer. Of course if you do that with ethernet switches (as do many metro networks) you quickly find out that Ethernet is a poor carrier protocol. PBT strips from Ethernet all the qualities which make it a poor carrier protocol and adds carrier fault management.

Best of all, there is a proposal to use GMPLS to provision the PBT service. The ease of service provisioning is one of the real advances of MPLS, so it is nice to see this retained.

Using Generalised MPLS is a good idea, as this allows a path to cross technologies, so customer links can cross the routed backbone (presumably in a VPN QoS class).

Notes on PBT

Transparent Ethernet bridging without flooding or learning. Topology if pre-configured, as is a fast-reroute path. 802.1Qah describes the Ethernet encapsulation.

A whole heap of standards for diagnosis.

802.3ah Ethernet in the First Mile describes the customer-facing port and its OAM.
802.1ag Ethernet connectivity fault management.
ITU-T Y.1731 gives fault management and SLA monitoring.

The fast re-route path is activated by ITU-T G.8031 Ethernet Protection Switching.

A ID for GMPLS to create the topology for the main and fast re-route path. This will eventually need IS-IS-TE to discover the topology.