Does portfast really help in STP convergence?

johnifanx98

Say a switch has two trunk ports and a bunch of end-station ports. Specify portfast to all end-station ports. However, for those end stations to forward data frames out of the swithch, they still have to wait for the trunk ports to converge. So, what is the point to use portfast here?

Mrock4

Without portfast, when you connect a host, the port generates a BPDU with the TCN flag set to the root bridge (in 802.1d), and the root bridge receives that, and sets the TC-flag on all of it's BPDU's, which cause the bridges in the network to flush their CAM tables.

In 802.1w, when a root or designated port goes to FWD state (except edge ports), root and designated ports (on the bridge that experienced the port transition) set the TC bit in their RST BPDU's.

In short, it actually does, but it doesn't seem like it from a high level view. Constantly having your bridges experience topology change results in an unstable network (at least in my eyes), so portfast helps to eliminate the unnecessary Topology Changes.

You also have to remember that those edge ports only need to wait for the truck to converge once- after that without any changes on the trunk, the host will plug up and transition to FWD, which will result in the host being online much quicker. Technically this transition is part of STP convergence.

networker050184

That would only happen in the scenario where the hosts boots at the same exact time as the trunk is coming up. Usually you are going to have the network up and stable before adding hosts.