[Cerowrt-devel] Next release plan (sugarland-2)
dave.taht at gmail.com
Wed Oct 3 21:18:13 EDT 2012
I like codenames, and cerowrt-sugarland thus far has been lucky, so...
This time around I plan to focus on the 3 packages everybody seems to
want, and keep hacking at qos-scripts and simple_qos.
minissdpd is done (in ceropackages) fixing upnpd looks easy, haven't
looked at minidlna yet.
Fiddling with qos effectively however requires I get some netem boxes
up to simulate long RTTs. While I have got the vyatta boxes to do so,
they are short on documentation, and I haven't even got a console
prompt out of them yet. Might acquire more/different boxes. Might get
the time and bum the hw to slam a video card into them to get them
updated. Don't know yet.
It would be nice to have a set of stable netperf servers around the
globe to use also, with well characterized behaviors.
I have a ton of data from the first field deployment attempt
http://www.bufferbloat.net/projects/cerowrt/wiki/Yurtlab/ - but unless
I have to give a talk about it I think that crunching it down can wait
until after we get 3.6 based versions of everything and a
repeat/refinement of the tests. The core things derived from that
experiment (memory use issues, a need for mfq_codel, early experiments
with driver buffering) are things we're fixing anyway.
Is anyone going to:
I applied for travel help with LF, didn't get any, may cancel the
trip. I am deeply grateful for the outpouring of donations and
subscription support - but so far 62% of the people that tried to
donated/subscribe last time haven't got the news of their
I'm still very encouraged. Certainly enough to be able to commit time,
look at our problems, and make a budget for cerowrt/bufferbloat.net
efforts for after my personal cashflow problem stops being so dire.
Try to come up with ways to expand it. Etc.
I wish I didn't have to stay even partially focused on the funding
problems, but I am going to stay on it until I'm sure we can meet some
reasonable goals on a regular basis.
I'm mostly happy to announce today that a certain very big ISP has
asked me to take a hard look at their network and see what can be
done. That contract should get me back to where I can eat regular by
the end of the year, and that contract does not conflict with the
cerowrt effort, except that I'll have less time for all this overall,
and phases where I'll have no time for it. More news on what will
happen there later.
So I hope to expand and parallelize the bufferbloat battle across more
people, and find ways to automate some things that will keep
incremental progress rolling along regular. Discussed adding cerowrt
to the openwrt buildbot system yesterday, as one example.
In late november, I hope to restart the campground field test and add
a new bloatlab on the east coast.
This month, I MAY "ping-pong" the release of the ubnt stuff I'm using
at the yurtlab as prep for sugarland-2, and I HOPE to get that ubnt
stuff done by oct 13th, as it's not gated on the same stuff cerowrt
Between now and mid-october:
I would like to set aside a half day towards getting those interested
in building cerowrt/openwrt up and running.
That's more than can possibly be done this month. By me, anyway.
Anybody else up to anything interesting?
I just learned that planetlab has linux 3.5 under test for a move to
std linux containerization.
Perhaps at last we can get some servers up for worldwide testing!
PS I have stuck a codel, fq_codel, ns2_codel, efq_codel enabled x86_64
Linux-3.6 kernel here:
Fixing bufferbloat with cerowrt: http://www.teklibre.com/cerowrt/subscribe.html
More information about the Cerowrt-devel