Hi,

On Sun, Jan 27, 2013 at 11:04:28PM +0000, Nick Hilliard wrote:
> On 25/01/2013 19:45, Gert Doering wrote:
> > Which is not god-given but a design decision by certain BUs that *should*
> > be able to get a faster CPU than a Z80 these days...
> 
> but where do you want to stop?  128 vlans means 128 customers with a single
> vlan each.  What happens when you want 2000 customers or 4094 customers?
> If you have a switch accepting hellos every 2 seconds per vlan and then
> running the spanning tree calculation for each of these every time a
> designated link flaps, you'll trash any cpu.  The protocol itself doesn't
> scale - it's great for small networks but it cannot scale beyond relatively
> small installations.

Oh, please.  Cisco can do it amazingly well on a 6500, which really does 
not have that much of a CPU.  It's a matter of having sane algorithms (STP
itself is not *that* complex to get right), a reasonable scheduler, and
a bit of beef in the CPU.

The pure number of VLANs is not the real deciding factor either - having
1000 VLANs on a single port each is less burden than 100 VLANs, but all
of them on 50 trunk ports.   (I could see a router vendor having a limit
there, like "your number of active STP port-vlans must not exceed 100.000"
or such, but "128 vlan limit" is just silly^W product marketing)

gert
-- 
USENET is *not* the non-clickable part of WWW!
                                                           //www.muc.de/~gert/
Gert Doering - Munich, Germany                             [email protected]
fax: +49-89-35655025                        [email protected]

Attachment: pgpbooiMVn732.pgp
Description: PGP signature

_______________________________________________
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