[Olsr-users] IEEE 802.11s

Juliusz Chroboczek (spam-protected)
Fri May 23 18:21:42 CEST 2008


> let's get this discussion over and finished please.

I'm sorry to disagree with you, but I like to think that most users of
OLSR are interested in mesh networking in general, and would want to
know more about alternate technologies.

> 1) YES, any properly 802.11s based device will have the possibility  
> to have olsr from olsr.org running on top of it (since it is layer 2  
> and olsr is layer 3)

Done naively, though, this is going to be horribly inefficient.  OLSR
uses link-layer multicast to flood its routing data, and 802.11s
implements multicast rather inefficiently.

Additionally, the link-quality sensing in OLSR is going to yield wrong
results over 802.11s (any sufficiently dense 802.11s subnet is going
to appear to have a cost of 1).

So running OLSR over an IEEE 802.11s network (as opposed to simply
exporting the 802.11s mesh as an HNA) is going to suck big time.

> and *end*this*discussion*. It is useless.

(giggle)

One thing to realise is that you should not see 802.11s as the enemy.
The main factor slowing down the adoption of mesh technology is the
centralised ESS model (bunch of wireless APs controlled by a central
server) promoted by Cisco and friends.  802.11s has a chance of making
mesh technology popular, which will prompt the move to layer-3 mesh
networking technologies as the limitations of layer-2 implementations
become apparent.

Don't take me wrong -- I believe that 802.11s will work great for
moderate networks.  However, 802.11s requires synchronised
time-stamps, and it uses a rather inefficient flooding protocol for
every broadcast/multicast frame (every ARP, every IPv6 ND, every DHCP
request, every zeroconf announcement).  Furthermore, 802.11s can only
integrate networks using the IEEE 802.2 frame format -- there's no
easy way to add, say, a PPP-over-GPRS link to your mesh.

For all of these reasons, I'm doubtful that 802.11s will be useful for
some kinds of deployments, such as the community networks many of us
are interested in.  People will try, though, they'll fail, and switch
to layer 3 technologies.

                                        Juliusz




More information about the Olsr-users mailing list