[Olsr-users] Very slow and Spotty Mesh

[Henning Rogge]

On Wed, Jun 29, 2016 at 12:05 AM, Ray Berotsfield
<(spam-protected)> wrote:
> Hello all,
>
> I have a new mesh consisting of about 20 nodes. 15 of them will be static
> (running Linux) and 5 of them will be mobile (running Windows) - moving
> about the mesh at about 3 m/s. To test the mesh's ability to handle the
> mobile nodes, I am using remote desktop to the Windows machine from the
> control station. The issue is, as soon as I am out of range of the mobile
> node, the connection becomes extremely slow, getting data very sporadically,
> and at some point it just utterly fails. I expected the data to start
> hopping over the nodes, but that was not the case. I have tried to tweak the
> HelloInterval and TCInterval in the config file because I thought that maybe
> the routing tables were not being updated fast enough to handle the speed of
> the nodes, but this did not improve things too much.
>
> I attached the config file for the Linux nodes (olsrd.conf). The IP is all
> that changes. Below is the Hello and TC message configuration I am using.
>
>   # Hello interval in seconds(float)
>     HelloInterval 1.0
>
>   # HELLO validity time
>     HelloValidityTime 5.0
>
>   # TC interval in seconds(float)
>     TcInterval 0.5
>
>   # TC validity time
>     TcValidityTime 5.0
>
>   # MID interval in seconds(float)
>     MidInterval 10.0
>
>   # MID validity time
>     MidValidityTime 30.0
>
>   # HNA interval in seconds(float)
>     HnaInterval 30.0
>
>   # HNA validity time
>     HnaValidityTime 600.0

Any reason why you run a shorter TC interval than the Hello interval?
This doesn't make much sense...

> I ran a test with three nodes, one Windows node(192.168.1.69), one Mac
> node(192.168.1.75), and one Linux(192.168.1.14). At close proximity, I was
> able to ping between nodes just fine. I was also able to remote desktop from
> the Mac into the Windows machine. I then moved the Linux node about 75 feet
> from the Mac, and the Windows 75 feet past it. As expected, data from the
> Windows node was hopping over the Linux node to get to the Mac (traceroute
> showed hop from 69 to 14 to 75. The remote desktop connection was still
> working OK, although a bit slow (even using high end wireless cards and
> antennas). After about a minute, it suddenly dropped and I could no longer
> remote desktop or ping the Windows machine. To regain connection I had to
> bring the Windows node back within wifi distance of the Mac. I then moved it
> back again 75 feet from the Linux to see if it the data would rehop over the
> Linux node like it had done before, but I had no luck getting it to connect
> again. Despite being far more advantageous to hop over the Linux node,
> "traceroute" showed that data from the Windows node was trying to go
> directly to the Mac without any hops. To see why this should not be the
> case, I attached a screenshot of my OLSR terminal session.
>
> The Mac and Linux are both running the latest release of OLSR (0.9.0.3) and
> the Windows is running an older version as I do not know how to compile the
> latest version on Windows (0.5.6).
> I am sorry if that is confusing. Please let me know if you need any more
> information.

First good advise, get rid of Windows... version 0.5.6 is somewhat
broken and horrible outdated.

> Here is a summary of my questions:
>
> 1. Can OLSR simply not handle slowly (slow jog speed) moving nodes? That
> would be very odd, given that this is a MANET protocol.

It can handle this...

> 2. Does my configuration look appropriate for the given setup? If not, what
> should it look like?

Start with an empty configuration just to see if everything is working.

> 3. I suppose the issue could be that the outdated version of OLSR running on
> the Windows could be causing this. But why would it work fine for a minute
> and then stop? If this is the issue, could someone give me some tips on how
> to compile the latest OLSR on Windows? That information would be very useful
> for the community.

Like I said above, 0.5.6 is ancient and has a horrible number of
subtile (and some not so subtile) problems.

> 4. This is slightly off-topic, but is there a way to make the Windows node
> not act as a relay, but as an endpoint that can move about the mesh. I don't
> really want it to act as a node for routing other traffic.

I don't think so. olsrd is quite limited in its handling of
Willingness and MPRs.

Henning Rogge