[c-nsp] general question on VRFs and FIBs...

[Gert Doering]

Hi,

On Tue, Sep 27, 2011 at 11:34:02AM +0200, Vitkovsky, Adam wrote:
> So is junos and ios really coded to carve up a separate FIB tables each 
> with it's own table preamble and stuff + adjacency table for each vrf? 

As you need a separate adjacency table anyway, what's wrong with that?

> As I believe it's possible to have a common FIB with some sort of a marker asociated with each table entry -describing the vrf participation

Think through the bits of my previous mail (which you quoted!) on why this 
is tricky, and makes either managing the FIB complicated (= bug-prone) or 
lookups slow.  You don't want either, so where's the percieved benefit of
having a single FIB "with markers"?

> Than wen you come out an interface asociated with a particular marker (vrf)
> the lookup would be done only on a subset of prefixes with the marker matching the one asociated with the interface 
> -though I'm not sure whether this would work with the common adjacency table and it's l2 overrides

Reasonably fast data structures don't permit "and only consult those
bits in the trie (or tree) with a marker matching" - because that wouldn't
be fast anymore.

gert
-- 
USENET is *not* the non-clickable part of WWW!
                                                           //www.muc.de/~gert/
Gert Doering - Munich, Germany                             gert at greenie.muc.de
fax: +49-89-35655025                        gert at net.informatik.tu-muenchen.de
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 305 bytes
Desc: not available
URL: <https://puck.nether.net/pipermail/cisco-nsp/attachments/20110927/4cbb12c5/attachment.pgp>