The difficulty is the words

Shortest Path Bridging is simpler than what it replaces. It does not feel that way on first contact, because seven unfamiliar terms arrive in one paragraph and nothing signals which are the same kind of thing.

Sort them into three piles and the technology stops resisting.

Transport: the backbone VLANs and the backbone MACs. Service: the , and the L2VSN or L3VSN it represents. Role: what a given switch does, edge or core.

The one fact that reorients everything

A B- is not a VLAN.

It does not flood unknown unicast, broadcast or multicast. It forwards only on backbone MAC tables that has provisioned from shortest-path trees.

There is no spanning tree. Nothing is blocked. Every link carries traffic, because the forwarding decision came from a computed topology rather than from watching frames arrive and guessing.

Once that lands, the rest follows: if forwarding is computed, the core does not need to learn anything, and if the core does not learn, it does not need to know your services exist.

The service identifier

The I-SID is a twenty-four-bit number carried in the 802.1ah backbone header. Twenty-four bits is about sixteen million, against a VLAN tag's four thousand — which is why a fabric can give every service its own identifier instead of rationing them.

An I-SID is not a VLAN and not a subnet. It becomes one or the other by what you bind it to:

  • Bind it to an edge VLAN and it is an L2VSN. Two sites configure the same I-SID against their local VLAN and share a broadcast domain, with nothing configured in between.
  • Bind it to a and it is an L3VSN, carrying routed traffic for one routing instance.

Many deployments encode the VLAN into the I-SID — 20010 for VLAN 10 — so the mapping reads at a glance. That is convention, not requirement.

Two backbone VLANs, and why

A fabric normally runs two B-VLANs, 4051 and 4052 by default on Extreme equipment. Services are distributed between them by the low bit of the I-SID: even on one, odd on the other. Two shortest-path trees, each carrying roughly half the services.

That distribution is the default rather than a rule of the standard. When one service behaves differently from its neighbours, which tree it is on is the first thing to check.

Edge and core

A backbone edge bridge is where a customer VLAN or VRF meets an I-SID, and where MAC-in-MAC encapsulation is added and removed. Services are provisioned here.

A backbone core bridge forwards on backbone MACs alone and never sees a customer .

That is the operational claim is actually making: adding a service means touching the two or more edge switches where it appears, and nothing in between. The core is untouched, not because of a design convention, but because it genuinely does not carry the information.

The same property sizes the core's forwarding table by the number of fabric nodes rather than by the number of end stations, which is why the encapsulation exists at all.