[Olsr-dev] Multipath routing in OLSRD
Hannes Gredler
(spam-protected)
Tue Apr 28 11:07:18 CEST 2009
extending the infrastructure and add a strawman (=ECMP routes only)
is perhaps a good idea anyway - later on we can decide what fancy algorithms we can
out on for determining path diversity, loopfreeness for non ECMP paths etc.
/hannes
Markus Kittenberger wrote:
> I would also suggest you share your ideas about multipath stategies here,..
> because if you just send data to all neighbours that would not send the
> packets directly back to you, you just multiply the chances to get
> routing loops on one of your path,..
> i mean, if you take the risk of having more problems, you should have a
> very good stategy to really gain better significant improves,..
> i mean if 1 or 2 hops later all you packets will go again through the
> same link, its for sure not worth the risk/effort.
> and if your plan is to send packets on every node to "every" posibble
> route, you may use quite long, bad routes, and suffer with enourmous
> packet loss/reordering problems (if you do not gurantee that "same"
> traffic takes (mostly) same route)
> i don`t want to say, multipath will not work, neither do i know your
> plans, but first step imho is to solve well some theoretical issues,
> before starting to hack this feature in,..
> regards Markus
>
> On Tue, Apr 28, 2009 at 9:58 AM, Henning Rogge <(spam-protected)
> <mailto:(spam-protected)>> wrote:
>
> Am Tuesday 28 April 2009 00:33:36 schrieb Adam Nowak:
> > Hello,
> > I'm new to this list, and to all the olsr stuff too. Sorry, if
> I'm asking
> > about well known issues. I'm interested in implementing multipath
> > extension to OLSRd. At first, I'd like to add some alternative
> routes to
> > the same destination (route entries which has the same
> destination but
> > other next-hops). As I've noticed in the code, there is the entry
> > "rt_best" which is responsible for selecting best route. I don't
> > understand, how OLSRd interacts with system routing table in
> > olsr_update_rib_routes() function, can you help me work it out or
> give
> > some advices?
> > Any help would be appreciated!
> The main problem you have to solve is how to force your packages to
> stay on
> the "secondary" route. You cannot do this "hop by hop" because the
> next hop
> will just send your packages to the shortest route again.
>
> For efficient multipath you need some "distance" (2 hops or more)
> between the
> two routes to prevent collisions between the two paths.
>
> Henning
>
> *************************************************
> Diplom Informatiker Henning Rogge
> Forschungsgesellschaft für
> Angewandte Naturwissenschaften e. V. (FGAN)
> Neuenahrer Str. 20, 53343 Wachtberg, Germany
> Tel.: 0049 (0)228 9435-961
> Fax: 0049 (0)228 9435-685
> E-Mail: (spam-protected) <mailto:(spam-protected)>
> Web: www.fgan.de <http://www.fgan.de>
> ************************************************
> Sitz der Gesellschaft: Bonn
> Registergericht: Amtsgericht Bonn VR 2530
> Vorstand: Prof. Dr. rer. nat. Maurus Tacke (komm. Vors.), Prof. Dr.-Ing.
> Joachim Ender (Stellv.)
>
>
> --
> Olsr-dev mailing list
> (spam-protected) <mailto:(spam-protected)>
> http://lists.olsr.org/mailman/listinfo/olsr-dev
>
>
>
More information about the Olsr-dev
mailing list