Date: Sat, 15 Nov 2003 04:17:32 +0300 (MSK) From: "."@babolo.ru To: Andre Oppermann <oppermann@pipeline.ch> Cc: David Gilbert <dgilbert@dclg.ca> Subject: Re: what about 5000 .. 10000 VLANs in one system? Message-ID: <1068859053.126816.95547.nullmailer@cicuta.babolo.ru> In-Reply-To: <3FB56251.F58DBF6A@pipeline.ch>
next in thread | previous in thread | raw e-mail | index | archive | help
> David Gilbert wrote: > > > > >>>>> "Brooks" == Brooks Davis <brooks@one-eyed-alien.net> writes: > > > > >> Why not simply index directly into an array of 4096 pointers? > > >> Anybody running that many VLANs can afford the extra 16 kB per > > >> physical interface. > > > > Brooks> I suggested the balanced tree because we've got two > > Brooks> implementations in sys/tree.h, but you are correct that the > > Brooks> space probably isn't worth the overhead of the trees. You'd > > Brooks> have to use per physical interface trees anyway, so that part > > Brooks> would be the same. > > > > I would vote for the 4096 pointer model (or at least an option for > > same). We often use machines with 100 or more vlans. Constant time > > packet delivery is a "good thing" (tm). > Yes, my vote would go there as well. theoretically each ethernet interface can have about 4000 VLANs, and 26 ethernets in one PC ip possible (but my routers maximum is 12) so table expandable by 2x is interesting (4096, 8192, 16384) Sorry my bad English
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?1068859053.126816.95547.nullmailer>