[Battlemesh] flent tests run?

[Ferry Huberts]

I'm happy that at least _some_ tests were run, so that we have at least 
something.

I'm eagerly awaiting the graphs. Unless someone shows me how to generate 
them myself obviously ;-)

Then, I have thought about the process and talked a bit with a few 
people and I'm convinced that we can come up with a way to bring up the 
test environment well ahead of time.

The idea we came up with is to do something like this:
- Use a simulated environment (like core, see 
http://www.nrl.navy.mil/itd/ncs/products/core) to bring up the wibed 
fully functional with the overlays in place that are validated by the 
protocol teams.
- Once that works, bring up the same setup but now on a small network of 
physical nodes, the ones that will be used during the WBM. This can be a 
very small network of only a few nodes. This is to validate that the 
whole thing will actually run on physical nodes.
- Finally, run it on a bigger network (if available) to work out any 
last wrinkles.

All of this (or at least most of it) can and should be done well ahead 
of the next WBM.

This should be feasible and it successful then we can use the setups to 
track the stable releases of all projects quite easily, so that there 
are no more surprises like a few WBMs ago.

As an extra bonus we can even (if we want) run the planned tests in a 
simulated environment to get an idea of what to expect and then later 
compare that to the actual tests run during the WBM.
And for the protocol developers, they might even want to use the results 
of the simulated environment for optimising their 
protocols/configurations, depending on how the tests are setup.


Thoughts?

On 09/05/16 13:54, Daniel Golle wrote:
> Hi!
>
> I also volunteered to run some flent tests on OLSR, however, I wasn't
> aware that this volunteer position ("We need a laptop for the far end
> of some tests!", me: "Sure, we can use mine.") would require me to
> understand and *select* the tests to run with flent, a graphical tool
> I have never used before and didn't perceive to be very stable (it
> crashed a lot, depending on the test chosen).
> Being on that far-away egde of the mesh, I managed to run two tests and
> went back to report and ask which tests to actually run. By that time,
> we were already wrapping up and decided to stop at this point :(
> The sad thing was imho that at this stage we had the testbed sorted out
> and all left was to run the actual tests and maybe some minor fixes for
> some protocols -- imho the main problem this year was to use a common
> overlay for ALL protocols which wasn't ready in time.
> Having an overlay for each protocol also has the layer 8+ advantage of
> the protocol teams really feeling responsible for 'their' overlay and
> testing to start once one of them is ready.
>
> Anyway, find the results of the two tests I run on OLSR attached.
>
>
> Cheers
>
>
> Daniel
>
>
>
> _______________________________________________
> Battlemesh mailing list
> Battlemesh at ml.ninux.org
> http://ml.ninux.org/mailman/listinfo/battlemesh
>

-- 
Ferry Huberts