Hi,
at the last Wireless Battle Mesh there were some nice graphs for the throughput and number of hops chosen - except for batman-adv because they used the layer 3 ping-tool for that.
We should probably look into having a graph about that for batman-adv at the next WBM, I think. Then we can compare with the other routing protocols via the throughput statistics whether batman-adv chooses too short routes and therefore.
For my current experiences with the number of hops, I can't say anything for sure yet. Years ago I had issues with batman-adv selecting too short routes in a specific indoor setup. But that turned out to be a mean hidden-node problem instead: batman-adv switched to a shorter - and unusable - route a few seconds after starting iperf.
From the Freifunk community mesh in Lübeck, I had been playing a
little with increasing the hop penalty on the VPN gateways a few weeks ago, with the intent to make the wifi-mesh more preferable compared to the VPN uplinks.
I wasn't able to get a workable route from this node: http://map.luebeck.freifunk.net/#!n:24a43c99ac30 to this node: http://map.luebeck.freifunk.net/#!n:647002d15c36
It'd take about six hops to get there. All links on this path are green, meaning they have a TQ above 200.
Unfortunately with the last hop the TQ dropped too low, the nodes aren't within each others horizon.
Also, we are still using batman-adv 2013.4.0 here, so we actually still have the hop penalty of 15.
Cheers, Linus
PS: Though I guess it doesn't make much sense to fiddle some more with the hop penalty now. The new throughput metric with BATMAN V is probably the better approach to mobile adhoc networking :).
On Tue, Mar 31, 2015 at 02:18:56PM +0200, Simon Wunderlich wrote:
Hi,
to cite from my original answer:
We did have some discussion on that patch too:
https://lists.open-mesh.org/pipermail/b.a.t.m.a.n/2014-June/012155.html
I wrote that patch because we had problems in commercial installations and dual radio access points, which were fixed by this patch (as described in the mail). I didn't hear any practical problems so far, you are the first one.
I guess its not easy to have a "one value fixes all problems" hop penalty, but its at least worth to discuss it - so thank you very much for your input.
Since we got this mail on the mailing list now, I'd like to hear opinions from other people and/or communities.
Thanks, Simon
On Monday 30 March 2015 21:27:26 Ruben Wisniewski wrote:
Forwarded private message, on request by Simon.
On Friday 27 March 2015 16:52:14 Ruben Wisniewski wrote:
Hello Simon,
we just updated our mesh, to the batman-adv-version 2014.4. Now your patch batman-adv: increase default hop penalty[1] is included.
Since we got very large mesh-networks, we have some troubles with your patch, which makes long (needed) pathes unusable, because they TQ drop to lower zero.
Else, we got some old and some new batman-adv versions, this change overloads the old nodes, because their TQ are better, and similiar links over new nodes got ignored.
Else I see no improvement overall, because the usual urban link-wide is about 30-50 meteres, and we got parts of the net where we need 4-5 hops to reach the last node. Your Patch seem to limit these pathes to something around 3 hops. So parts of our net are now offline, and the speed overall doesn't seem to be improved.
After these experience I like to ask you, if you mind reverting this patch again. Else we have to work around with resetting to 15 again on boot, which I think is not the prefered solution.
[1] http://www.open-mesh.org/projects/batman-adv/repository/revisions/7644650b b7 66bd4c7b6be5d97e6b1c3ed93a38d7
Best regards,
Ruben