[Battlemesh] battlemesh v9.1

[Dave Taht]

One thing I would like to have is the referenced flent tests (that did
complete for bmx) conducted, for reference, against the existing wibed
setup, in the lab (wherever it exists). The results at battlemesh were
predictably horrific.

http://blog.cerowrt.org/flent/battlemeshv9/t1/bmx_fail.svg

http://blog.cerowrt.org/flent/battlemeshv9/t2/blech.svg

and seeing those duplicated in someone elses lab would be good. (?)

We are making great progress in finally ripping out all the queuing
delay in wifi:

http://blog.cerowrt.org/post/rtt_fair_on_wifi/

as well as scheduling delay:

100 stations, 1 flow each before: http://blog.cerowrt.org/flent/drr/10tothe5.svg
after: http://blog.cerowrt.org/flent/drr/newcode.svg

and I hope that by the next battlemesh, (and preferably before, in
labs around the world), we'll have got much of this problem licked.

It makes little sense to be measuring protocol convergence when the
underlying network cannot actually handle real traffic.

The progress of this research is being blogged at:
http://blog.cerowrt.org/post/ with references to the source patches if
anyone wants to join in (ath10k only at the moment, hope to add ath9k
and mt72 soon).

And *all the data* for all the experiments is available from the
github repository: https://github.com/dtaht/blog-cerowrt