[rbridge] per-VLAN instances of IS-IS
anoop at brocade.com
Fri Jun 15 13:51:48 PDT 2007
> -----Original Message-----
> From: Eric Gray (LO/EUS) [mailto:eric.gray at ericsson.com]
> Sent: Friday, June 15, 2007 1:05 PM
> To: Anoop Ghanwani
> Cc: rbridge at postel.org; Radia Perlman
> Subject: RE: [rbridge] per-VLAN instances of IS-IS
> There are two issues you're attempting to address in
> this mail. One is the issue with advertising, and you have
> further broken that into unicast and multicast. The other is
> with how the RBridge link state protocols scale.
> On the first issue, you agree that having the default
> mode for acquiring unicast MAC reachability be via learning,
> addresses that portion of the scalability issue. For
> multicast, you seem to assume that multicast optimization is
> required - because the only reason why acquiring multicast
> address reachability would be needed is if you absolutely
> require multicast optimization (instead of treating multicast
> as broadcast - as is still allowed for a standard bridge).
All I can say is - try selling a bridge that doesn't do
IGMP snooping to someone that actually uses multicast.
That will address your concern on why the optimization
is absolutely essential.
> On the issue of link-state protocol scalability, do you
> have any evidence to support your claims for non-scalability?
> Does it (your evidence) include a comparison of scalability
> of multiple instances to scalability of similar single
> instance solution?
> Throwing the number of adjacencies around does not - in and
> of itself - prove much of anything.
I don't know too many midrange routers that are
capable of supporting 4K adjacencies. I do know
several midrange switches that are perfectly
capable of supporting all 4K VLANs and mapping
them to MST instances.
Why don't you prove that it scales, instead of
having me prove otherwise? :-)
More information about the rbridge