On Mon, 13 Apr 2020 at 00:44, Mark Tinka <[email protected]> wrote:

> What we've generally done is make sure the lowest level protocol reacts
> as quickly as possible. This way, it's (quick) reaction would inform
> lateral or upper-layer protocols to react accordingly.
>
> So we tune IS-IS to react quickly, and enable BFD just for it.

+1, only ISIS (and other SPT protocols) are topoogy aware. If ISIS
converges fast, and the rest of the protocol stack is correct, then
everything converges fast. LFA, rLFA, PIC, edgePIC and you can locally
converge once you are aware of the outage, ~immediately. That is,
convergence can happen in linecard HW, as the backup path is already
programmed there, the only thing it needs to know is if-down. And the
rest of the network not converged yet, are protected by the node which
is aware of the failure.

In JNPR this is accomplished by ECMP mechanism, a backup path is added
to LC HW ECMP with inferior weight, so it's not actually used. So if a
linecard can for any reason remove the superior weight (int down) it
can also start using the inferior weight, without understanding
anything about the fault.

-- 
  ++ytti
_______________________________________________
cisco-nsp mailing list  [email protected]
https://puck.nether.net/mailman/listinfo/cisco-nsp
archive at http://puck.nether.net/pipermail/cisco-nsp/

Reply via email to