[rbridge] TRILL over IP
radiaperlman at gmail.com
Sun Nov 20 23:00:01 PST 2011
I believe that Margaret's draft is a way of sending
already-TRILL-encapsulated traffic over an RB-to-RB link that just
happens to be IP rather than Ethernet. So there's no issue about
Margaret's draft can be used to interconnect two distant data centers
by glueing them together by having at least one RBridge, R1, in one
center, creating a TRILL link to another RBridge RB2, in the other
center. Once this is done, this becomes a single TRILL fabric. Other
than RB1 and RB2, the other TRILL RBs don't care that the link between
R1 and R2 is IP or Ethernet. It's just a link reported in the TRILL
I suppose if there were a lot of RBridges interconnected with a
backbone IP network, then instead of modeling the links across the
backbone as a bunch of point-to-point TRILL links, one might want to
consider the IP backbone as a single shared link, elect a DRB, create
a pseudonode, etc. This adds complications like wanting to multicast
Hellos, LSPs, CSNPs, etc. But in most cases, just reporting the
pt-to-pt links between pairs of RBs that are configured to create IP
tunnels to each other, and unicasting all the control traffic between
the pairs of RBs that want to mate across the IP backbone seems simple
and useful and sufficient.
Unless I'm confused about the problem being solved, or something, of course.
On Wed, Nov 16, 2011 at 9:13 PM, Margaret Wasserman <mrw at lilacglade.org> wrote:
> On Nov 16, 2011, at 11:38 PM, Joe Touch wrote:
> A packet comes into TRILL campus A. In the process f transiting that campus,
> it goes over an L2 link that ends up being encapsulated over L3, which then
> gets encapsulated over L2, and ends up arriving at that campus again.
> As TRILL over IP is currently defined, it is used as a link between two
> RBridges within a single TRILL campus, not as a way to connect separate
> TRILL campuses...
> Does that make a difference here?
> rbridge mailing list
> rbridge at postel.org
More information about the rbridge