[c-nsp] Maximum spannig tree instances

Tomas Daniska tomas at soitron.com
Wed Jul 15 09:30:45 EDT 2009


> 
> On Tue, Jul 14, 2009 at 3:45 AM, <A.L.M.Buxey at lboro.ac.uk> wrote:
> > Hi,
> >
> >> ... but it doesn't say anything about the number of STP instances.
> >
> > things go wonky when you have more than 1800 virtualports per slot
> > (which you didnt quite reach) (1200 on older eg 100mbit blades)
> > with 13,000 in total (PVST+), 10,000 in total (RPVST+)
> >

As a matter of coincidence, I've been in talks recently with our local
Cisco SEs for some 6k5/3750E design, mostly discussing RSTP vs MST. I
have asked about the 1800 virtual ports per blade limit and they say
this only applies to 61xx and 63xx cards - the 65xx and 67xx have no
such limit. There is a ddts that a message errorneously warning of
exceeding 1800 virtual ports on a 67xx is removed since SXI (or SXI1 it
was).



Re MST vs RSTP... the worst case in MST for us is that once you get any
tiny irregularity on a port, it gets to interoperability mode, which
means the port is calculated against CIST (MST0). And then, any issue or
TCN you have, everything gets propagated to all remaining instances,
causing MAC table flushes and other nice stuff for the *whole*
infrastructure.

We had an idea of having two independent MST domains interconnected by a
(VSS/Multichassis Etherchannel) trunk, so we could have STP events
contained within a single physical location. But with respect to
abovewritten the trunk would be in the interop mode, amplyfing all
events instead of separating the domains. We could have had BPDU filter
to solve this on the trunk, but obviously would lose loop prevention
because of that.

And not speaking of MST experience we had building a large-scale Metro
Ethernet network, with many access rings. We have repeatedly seen BPDUs
transported via EoMPLS pseudowires in 3750ME based rings causing NNI
trunks (running MST) get into P2P Edge mode and thus bringing the whole
ring down. Yes, this is more due to the pretty weird MPLS implementation
on the 3750ME, but nicely showing MST weaknesses...


So far, MST hase become a no-go for us unless there's a *very* strong
scaling requirement.


--

deejay
 

__________ Informacia od ESET NOD32 Antivirus, verzia databazy 4240
(20090713) __________

Tuto spravu preveril ESET NOD32 Antivirus.

http://www.eset.sk
 


More information about the cisco-nsp mailing list