[OLSR-users] WLAN and Broadcasts (e.g. OLSR)

[Nikolaos Tsarmpopoulos]

Hi,

I can confirm that this was a major issue on our network in Volos, 
Greece (http://vmesh.inf.uth.gr/). Specifically, when there was a file 
transfer between two nodes, broadcast packets used for HELLO messages 
around these nodes were lost (due to collisions, I suppose) resulting in 
unstable connections.

Also, I recall that someone suggested in the list the solution of 
broadcasting + unicasting of HELLO messages, where broadcasting aims at 
detecting new nodes, while unicasting aims at checking if known 
neighbors are still reachable. I haven't checked latest features of 
olsrd, but I think it is a planned feature for the new 
(non-OLSR-compliant) version of the protocol.

Regards,

Nikos

Sven-Ola Tuecke wrote:
> Joerg,
>
> some Wifi cards have 2 settings for speed and mode. Unicast rate and 
> multi(broad)cast Rate. If may have for example a multicast rate of 54g 
> while using unicast 1b. Look out for iwprivs on that issue.
>
> HTH
> Sven-Ola
>
> "Joerg Pommnitz" <(spam-protected)> schrieb im Newsbeitrag 
> news:(spam-protected)
>> Hello all,
>> I have stumbled accross a problem with olsrd. It seems that 
>> HELLO-Messages are only visible over a very short distance. Other 
>> protocols (ICMP-Echo, ssh, iperf) work fine, however. The difference 
>> is, that these protocols are unicast while olsr HELLO-messages are 
>> sent to the broadcast address of the subnet. I have verified that a 
>> broadcast ping has the same problems, so its not really OLSR-specific.
>>
>> This got me thinking. My understanding is, that 802.11 modifies the 
>> over-the-air coding of the data packets according to the error rate 
>> when talking to the peer. It trades bandwidth for reliability when 
>> the error rate goes up. However, this is a point-to-point concept. In 
>> AdHoc-Mode where you talk to all the nodes in a cell, this is not 
>> applicable. My suspiction is, that the 802.11 card sends the 
>> broadcast packets in whatever coding is the default for the current 
>> mode which might not be optimal for nodes that are further away. Has 
>> anybody seen something like this? Am I on the right track? Is there a 
>> solution? Something I could try?
>>
>> --  Thanks in advance
>>
>>       Joerg
>>
>>
>> _______________________________________________
>> olsr-users mailing list
>> (spam-protected)
>> https://www.olsr.org/mailman/listinfo/olsr-users 
>
>
> _______________________________________________
> olsr-users mailing list
> (spam-protected)
> https://www.olsr.org/mailman/listinfo/olsr-users
>