[rbridge] Range of appllicability (was Re: TTL only - was RE: New fields in shim header?)
caitlinb at broadcom.com
Fri Oct 13 09:43:03 PDT 2006
rbridge-bounces at postel.org wrote:
> Why not ask the question the other way? If Rbridges exist,
> why would anyone want to use an old fashioned STP or RSTP
> based bridge? :-) Simplifying a lot, Rbridges make better
> use of available links while various forms of spanning tree
> work by turning off and thus wasting links.
> Of course there are some reasons, like Rbridges probably
> require greater computational resources and bandwidth for
> routing messages. But as computation gets cheaper and link
> bandwidth keeps rising, these factors become less significant...
> While I suppose this is all an interesting discussion, I
> personally don't think it actually has much effect on the
> Rbridge specification. As long as rbridges are to be
> applicable where high end bridges are currently used, then
> they need to be able to support VLANs, optimize multicast,
> etc. Whether or not they are beneficial where mid or lower
> range bridges are used seems less important.
At the minimum there is the need for phased deployment.
I doubt there are many sites that would agree to bring
the entire subnet down so that every existing bridge
can be replaced instantly.
RBRidges can also make sense to provide a large campus wide
network connecting many departments/floors/rooms (not just
racks in a data center). In a campus deployment I suspect
that conventional Bridges will be around at least as long
More information about the rbridge