[OLSR-users] Host vs. Network route propagation with OLSR
Thu Jul 20 21:58:50 CEST 2006
I understand your confusion!
In an ad-hoc network _every_ node is a router, but the nodes are usually
part of the same subnet. Aggregated routing is based on the assumption
that a subnet is reachable on a network segment(on-the-wire). The problem
with MANETs is that all nodes are not reachable by one layer2-hop and
therefore host routing is nessecarry.
So the assumption from traditional routing that we can add an aggregated
route to a routers subnet fails. However, it is possible to build
something similar to what you describe using HNA(host/network
association). Nodes can announce external connectivity as subnets by
sending HNA messages.
On a sidenote I've been playing with the idea of making a olsrd version
that only uses propagated routes. Say you have a network consisting of
only the good ol' WRTs. Then every WRT is assigned a small subnet. The WRT
then uses normal ethernet/WLAN operation towards its clients(DHCP etc) but
does not (need to) do any NAT.
In such a network only aggregated routes would be needed in the MANET as
well. In addittion you would probably want a mechanism to build host
routes for maintainenc etc. but that wouldn't be too hard.
Ofcause then you would turn olsrd in to a "backbone routing protocol" but
so what :) Hosts could still connect if they were assigned a subnet upon
joining the network.
Anyways, I hope the explanation made senese (even is my last
thinking-out-loud section might not have) :)
> I'm new to OLSR and am finding the learning curve steep. I'd like to
> make several comments to clarify things in my mind and I'd be
> terribly grateful to hear your comments and corrections.
> It appears to me that there is a huge misconception between the
> designers of OLSR and much of the current user base (myself included,
> actually maybe I'm the only one with the misconception).
> As best I can tell, OLSR is fundamentally different than OSPF or RIP
> in that it propagates routes for HOSTS rather than networks - by
> default. I suppose this makes sense as many ad-hoc mesh networks may
> be comprised of "sensor nodes" in which each node both "does
> something" (measures the temperature maybe) and is also a router. In
> this way, each sensor is a possible path for data transmission, and
> since each node is a router, one need only route directly to
> interfaces and not to networks.
> However, this is very different from the model that many people are
> used to in which an ad-hoc mesh of nodes provide routing services
> between networks to which other clients can connect. For example, a
> community wireless project in which each node in a network provides
> both a local AP functionality for client computers and a wireless
> link to other nodes. In this scenario OLSR might be chosen so that
> non-persistent nodes installed on busses or ferries or whatever are
> handled gracefully. Here, OLSR is meant to provide routing between
> NETWORKS rather than just the network interfaces of the node. This is
> much more like a mobile meshed LAN providing dynamic networking for
> client machines rather than a mesh sensor network.
> When you set up OLSR and configure the "interfaces" entry in
> olrsd.conf, you propagate routes to those interfaces through the
> network. But you don't propagate routes to the networks those
> interfaces are attached to, and so clients on those networks are
> unreachable. This fact was and is very confusing to me, as a very
> similar entry in ospfd.conf will propagate routes to the networks
> those interfaces are connected to, rather than just the interface
> itself. A review of this list archives seems to show that it is
> confusing to many others too.
> I'm probably missing something, but I cannot see a reason why an
> "interface" entry in olrsd.conf should not simply propagate a network
> address rather than a host address. It seems to me the more general
> case of a network address would serve to provide a route to both the
> network interface and other addresses on that network - in much the
> same way as other common routing protocols.
> Thanks for listening - this wasn't meant to be a rant. I look forward
> to your comments.
> Val Schmidt
> University of New Hampshire
> Chase Ocean Engineering Lab
> 24 Colovos Road
> Durham, NH 03824
> e: vschmidt [AT] ccom.unh.edu
> m: 614.286.3726
> olsr-users mailing list
More information about the Olsr-users