[c-nsp] backup with STP

[Andrew Fort]

On 30/09/2005, at 7:46 AM, David J. Hughes wrote:

> I've been trying to delve back into memory to remember all the
> intricate bits and pieces of MST (it's been quite a while).  In the
> context of the original question, one of the driving points for the
> posters decision making was support for legacy dot1D devices.  Now,  
> you
> guys are far more familiar with MST than I am these days so I'm sure
> you can clarify this point.
>
>  From my memory, all dot1D devices were held in a "virtual region"
> outside MST's IST instance.  In this situation is a dot1D BPDU
> forwarded into the IST and as such broadcasted to all STP instances or
> is there another STP instance outside the IST for the dot1D devices?
>
> Like I said, it's been a while but I thought legacy dot1D support  
> was a
> weak point in MST.  RPVST+ naturally doesn't suffer from this sort of
> problem.

The IST does 'take care of things', but you do have to consider the  
implications.  If one connects the MSTP region with dot1d devices,  
it'll still work and links will be reliably blocked, but it'll be  
slow as all hell, and you don't want 3550s being the roots of the IST  
in this case (we find them causing longer temporal loops upon  
transitions).

The simple answer is don't use STP, but use RSTP, or if you need  
instances for better network utilisation, use MSTP.  If using MSTP, I  
highly recommend pre-provisioning all VLANs to instances, to avoid a  
really dumb operational issue you'll come across (a bunch of  
reconvergences during instance mapping reconfiguration).

To answer the original question, Cisco kit running MSTP will talk to  
other cisco kit running RSTP, PVST+, RPVST+ and STP.  I wouldn't  
engineer that as a requirement into my network, though.

Agreed in a network running PVST+ it would've been far simpler to  
transition to RPVST+, and harder to transition to MSTP/RSTP.  Such an  
original design choice of PVST+ may have been more based upon it  
being the only ball-game in town to load-balance, at the time.

> David
> ...
>

-andrew