[Olsr-dev] Thoughts about an ETX-FF-Metric plugin
Henning Rogge
(spam-protected)
Mon Jun 16 09:16:28 CEST 2008
At the moment we have problems (ETX values are not as stable as they were
before) between the new olsrd version (0.5.6-rc) and the old ones.
This problem has two reasons:
- we changed the ETX calculation to a pure Hello based system (as described in
the original MIT paper), the old one is based on packet sequence numbers
which use Hellos and TCs.
- the old olsrd agents have an instable timer implementation, so the Hellos
the new implementation depends on are sometimes late (sometimes 80% later).
The problem of the old ETX implementation was that it has an instable timing.
The number of routing packet transmitted over a link depends on the size of
the network. Combined with a fixed window size the rate of ETX-change is
dependant on the net size too, which is a bad idea in my oppinion.
This weekend I had some time to think about a "compromise" that might solve
our problem. I will try to summarize the algorithm:
- the routing plugin will count the total number of packet and the missing
packets (calculated by packet sequence numbers, which have NO duplicates !).
- Once every second the plugin calculates the ratio between captured and lost
packages and use this value to update the LQ of the link.
I see two advantages of this strategy:
1.) we can use all olsr packets on a link to measure it's quality
2.) the rate of ETX-change of a link is independant of the amount of routing
messages transmitted over a link.
What are you thinking about this idea ?
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)
Web: www.fgan.de
************************************************
Sitz der Gesellschaft: Bonn
Registergericht: Amtsgericht Bonn VR 2530
Vorstand: Dr. rer. nat. Ralf Dornhaus (Vors.), Prof. Dr. Joachim Ender
(Stellv.)
More information about the Olsr-dev
mailing list