[olsr-dev] OLSRd - traffic problem

[Marco Tidow]

On Tue, Jul.11. 10:11 -0400, Bastian Bittorf wrote:
> problem: OLSRd does a lot of traffic by "ambient noise".
> the DSL-lines with small _upload_ are overloaded. why is
> there so much traffic? we are talking about 5-7 kbytes/s.

Hi Bastian,

got this setup here (in Berlin):

Interface "tun0"
{   
    Ip4Broadcast                104.130.1.21
    HelloInterval    5.0
    HelloValidityTime   90.0
    TcInterval        15.0
    TcValidityTime      270.0
    MidInterval         15.0
    MidValidityTime     390.0
    HnaInterval         15.0
    HnaValidityTime     90.0
    Weight 2
}   

with arround 200 nodes at the moment,

   ||| seagull:~# route -n |wc -l
   ||| 237

it gives the following 'idle' traffic amount:


   ||| seagull:~# sh log.traffic.sh 3 'ppp0' 1
   ||| STAMP                              ppp0                 
   |||                                            RX         TX
   ||| 1152630100 20060711 17:01:40              998        942
   ||| 1152630103 20060711 17:01:43              998        820
   ||| 1152630106 20060711 17:01:46             1954       1354
   ||| 1152630109 20060711 17:01:49             1664       1942
   ||| 1152630112 20060711 17:01:52             2249       1640
   ||| 1152630115 20060711 17:01:55             1396       1634
   ||| 1152630118 20060711 17:01:58             2577       2109
   ||| 1152630121 20060711 17:02:01             1430       1068
   ||| 1152630124 20060711 17:02:04             2534       2360
   ||| 1152630127 20060711 17:02:07             1980       1483
   ||| 1152630130 20060711 17:02:10             2241       2806
   ||| 1152630133 20060711 17:02:13             1870        776
   ||| 1152630136 20060711 17:02:16             1892       2318
   ||| 1152630140 20060711 17:02:20              900       1445
   ||| 1152630143 20060711 17:02:23             1136        858
   ||| 1152630146 20060711 17:02:26             2514       2431
   ||| 1152630149 20060711 17:02:29             2461       1504
   ||| 1152630152 20060711 17:02:32             1602       2302
   ||| ...

seems to me, by slowing down the intervals, but keeping validity times
according to those used w/ wireless-interfaces, it should be possible to
stay w/ only one instance of olsrd in your setup.

propagation of links changing their LQ is of course much slower via the
tunnel this way.  since I´m using tunnels to access islands not
reachable via air only, there´s no garantee wether this config may produce routing-
loops.  on the other hand, slowing down the frequence of wired TC-mesg´s could be
compared to "packet-loss" of TC´s via air (IMO one of the reasons setting the
TcInterval usually _much_ shorter on wifi-links, typically 2-4s here).

looks like having inconsistent validity-times is more a concern, a ka if
nodes switch their routes at _very_ different times, noticeably to the user as
delays caused by routing loops. 

N.B.  the increased MidValidityTime >> TcValidityTime has proven to produce _a_lot_
     more stable routes than before, using the wrong relation
       MidValidityTime == HnaValidityTime << TcValidityTime;
     this is simply due to the fact, valid MID-info is required to interpret TC-
     mesg´s correctly, and - btw - MID-config does not change very often at all,
     ;-)

     (MID : multiple-identity, nodes having more than one interface/IP-addr are
            treated "as one hop" by olsrd)


- or - hm,

 is the increased traffic the result of having parallel
 wifi-links also ???  what happens if you shut-down the wifi-dev?


marco