[olsr-dev] Question about LQ theory

[Bruno Randolf]

On Tuesday 31 May 2005 14:57, Thomas Lopatic wrote:

> [802.11 broadcasts are sent at a lower bitrate than unicasts; the packet
> loss for broadcast packets is thus different from the packet loss for
> unicasts; hence, one might want to send unicast probe packets to measure
> unicast packet loss]
>
> Yes, true, you could also try and send extra unicast packets to each
> neighbour. However, as each node would send packets to each of its
> neighbours, that could be a lot of packets. Hence, we decided to stick
> to packets that were already there, i.e. OLSR broadcast packets.

the problem with 802.11 unicast/broadcast rates is at one osi layer beyond the 
one olsrd operates. olsr operates at the ip layer (layer3) and with different 
media (might be aswell ethernet, infrared, ppp, whatever) and the problem is 
at the wlan link layer (layer2). it is also called "communication gray zone" 
in some papers.

one solution for this problem is to configure the wlan driver so it sends 
broadcast packets with the same rate as unicast packets. this can be done by 
setting it to a fixed rate and/or by adjusting the basic rates (used for 
broadcast and management frames), depending on the driver.

> What we observe is a bit different anyway. Sometimes for some reason the
> measured packet loss is higher than the actual packet loss. I've read
> somewhere that broadcast packets do not get acknowledged in 802.11 and
> hence not re-sent in case of packet loss, which might be an explanation
> for this behaviour.

actually this is the whole point of the operation ;)

by measuring the packet loss of broadcast packets (which are not acked and not 
re-sent) we estimate how many *link-layer* retransmissions are needed for 
unicast packets which will be re-sent at the link layer (and maybe once again 
by the ip layer).

bruno

-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: not available
URL: <http://lists.olsr.org/pipermail/olsr-dev/attachments/20050531/6c15c031/attachment.sig>