[c-nsp] Divide large PVST domain?

[Andriy Bilous]

In most of the cases you won't notice any disruptions as long as your
(copper/fibre) links are up and running. STP timers are overly pessimistic
regarding propagation delay for the worst case when you need to communicate
topology change from the furthest right bridge to the furthest left in the
slowest possible medium. They're utterly optimistic about processing delay
though - it's easy to clog CPU effectively disabling BPDUs processing and
forwarding plane will happily switch frames based on outdated state. Your
radio links don't make it easier. I'd recommend you to switch to rapid
PVST+ at least which as previously mentioned doesn't care about diameter as
it doesn't rely on root bridge for topology changes notification.


On Wed, Jul 9, 2014 at 6:14 PM, Victor Sudakov <vas at mpeks.tomsk.su> wrote:

> John Gaffney wrote:
> > You could consider MST to reduce the number of STP instances.
> > Sometimes that can increase efficiency.
>
> Am I correct to assume that every time I need to move a vlan from one
> MST instance to another, my whole MST domain will fall apart until the
> MST reconfiguration is complete on all the switches?
>
> Somehow I don't like this idea.
>
>
> > However what you are probably looking for is "spanning-tree
> > diameter" command. If I recall that can make some auto adjustments
> > to STP timers to accommodate for large switched networks.
>
> Even with the timer adjustments the maximum diameter is about 18
> devices. Still my train does miraculously work provided the root
> switch is in the middle.
>
>
> > You should also be aware that 20 switches connected in a straight
> > line (hence middle switch) gives you a single point of failure.
>
> It is not exactly so because the radio equipment has a way to shunt
> the failed switch. So the continuity of the train can be quickly
> restored. But it is a different topic.
>
> I also have a backup root close to the middle of the train.
>
> This is also the reason why going all L3 is not an option. You cannot
> simply shunt a router.
>
> > If it is truly a "train" of switches you could also not run STP - no
> > ring/redundant link = no loop.
>
> As I have already said before, 'I have a "train" topology with some
> redundant links between neighboring switches'.
>
> There are however no rings in the sense that redundant links are only
> between neighbors.
>
>
> --
> Victor Sudakov,  VAS4-RIPE, VAS47-RIPN
> sip:sudakov at sibptus.tomsk.ru
> _______________________________________________
> cisco-nsp mailing list  cisco-nsp at puck.nether.net
> https://puck.nether.net/mailman/listinfo/cisco-nsp
> archive at http://puck.nether.net/pipermail/cisco-nsp/
>