Permalink
Switch branches/tags
Nothing to show
Commits on Mar 18, 2013
  1. Clean up whitespace.

    tohojo committed Mar 18, 2013
  2. Don't touch the 'lo' interface.

    tohojo committed Mar 18, 2013
Commits on Mar 17, 2013
  1. Fix bug w/limit on gbit

    tohojo committed Mar 17, 2013
Commits on Dec 12, 2012
  1. Updated debloat and debloat.sh slightly

    Dave Täht committed Dec 12, 2012
    debloat does not set the bql limit at 10Mbit as low as debloat.sh now
    does.
    
    Still need to detect bridges
Commits on Dec 3, 2012
  1. Added all kinds of data

    Dave Täht committed Dec 3, 2012
Commits on Nov 6, 2012
  1. Fixed unkind comment in the text.

    Dave Täht committed Nov 6, 2012
    A really huge problem is that most of our tests test for the long term
    behavior of tcp, when the real behavior is dozens or hundreds of short
    sharded TCP connections and dns lookups. Looking REALLY hard at that,
    is what is really needed in this test....
  2. Saved the latest draft of the rrul test.

    Dave Täht committed Nov 6, 2012
Commits on Nov 3, 2012
  1. Moved the low latency setting

    Dave Täht committed Nov 3, 2012
    Older devices really do need the TSO/GSO offloads at gigE speeds.
    
    Still, I wish we could turn them off if we knew when we could,
    better. I dislike GSO, in particular.
Commits on Nov 1, 2012
  1. more fixes

    Dave Täht committed Nov 1, 2012
  2. eliminated lua-ism

    Dave Täht committed Nov 1, 2012
  3. added description and copyright

    Dave Täht committed Nov 1, 2012
  4. bug: MQ setup is in hexadecimal, not decimal

    Dave Täht committed Nov 1, 2012
    switched to using a printf %x
  5. abstract fq_codel into QDISC

    Dave Täht committed Nov 1, 2012
    There are multiple variants of fq_codel in testing right now
    (only one in mainline). This makes it easier to try those with
    this script.
  6. Add setting lower BQL limit at 101Mbit and below

    Dave Täht committed Nov 1, 2012
    2 big packets is all you need. BQL usually autoconfigs with a value
    2.5x what is useful at lower speeds.
Commits on Oct 31, 2012
  1. started cleaning up debloat script

    Dave Täht committed Oct 31, 2012
    debloat just grew too big and too complex to live with anymore.
    It scares me, just to read it.
Commits on Oct 30, 2012
  1. made default QMODEL fq_codel_ll

    Dave Täht committed Oct 30, 2012
    pushing back on the quantum a little helps with bidirectional flows
    in asymmetric scenarios.
Commits on Sep 26, 2012
  1. debloat: catch up to cerowrt debloat

    Dave Täht committed Sep 26, 2012
    this adds support for nfq and efq_codel and related
Commits on Sep 19, 2012
  1. moved to a saner name

    Dave Täht committed Sep 19, 2012
  2. Whole bunch of benchmarks

    Dave Täht committed Sep 19, 2012
Commits on Aug 28, 2012
  1. Various forms of tests for latest codel

    Dave Täht committed Aug 28, 2012
Commits on Jul 11, 2012
  1. lupin related hwtests

    Dave Täht committed Jul 11, 2012
    I started at automating a string of tests to run nightly across
    multiple servers across multiple paths. Natually I ended up with
    far more variables than I'd planned for so I have to rethink
    the test framework.
    
    The nice thing is that these tests already thoroughly exposed
    a major bug in VI/VO queue handling that is now fixed in openwrt mainline.
    
    http://www.bufferbloat.net/issues/401
    
    Which I have to get out to the 12 machines in the testbed...
Commits on May 16, 2012
Commits on May 10, 2012
  1. Added codel to debloat script

    Dave Täht committed May 10, 2012
  2. Switch simple_qos to codel

    Dave Täht committed May 10, 2012
Commits on Apr 28, 2012
  1. Better ingress filtering and classification.

    Dave Taht committed Apr 28, 2012
    Now we have full diffserv based ingress classification.
    
    Sadly, along the way, I realized that iptables wasn't useful on ingress
    classification. Or I'm doing it wrong.
    
    Ah well, if traffic is classified at the edge, this helps.
Commits on Apr 23, 2012
  1. IFB doesn't obey iptables. And iptables loses interface at postrouting

    Dave Taht committed Apr 23, 2012
    So I fixed the postrouting stage. As for the ifb fix...
    
    ... I'm beginning to understand imq.
  2. Bumped up red min and max for ifb

    Dave Taht committed Apr 23, 2012
    At 20Mbit down it was hitting it too hard. This is too much
    (by a lot), but it doesn't matter so much here. Right now,
    anyway.
Commits on Apr 22, 2012
  1. Integrated simple AQM script for up/downlinks

    Dave Taht committed Apr 22, 2012
Commits on Apr 16, 2012
  1. insert needed kernel modules

    Dave Taht committed Apr 16, 2012
  2. Simple_qos and the initial ingress model 2Mbit/24Mbit

    Dave Taht committed Apr 16, 2012
    Shoots the elephants rather well.
Commits on Apr 15, 2012
  1. cdf plot of ping and netperf data

    Dave Taht committed Apr 15, 2012
  2. Formalized formally ad-hoc test for exercising shaper bins

    Dave Taht committed Apr 15, 2012
    Doing a comprehensive open source test suite is needed work.
    Doing one well will take much work.
    I'd hoped someone else would do the work.
    
    I have a pile of scripts like this that can be gradually
    generalized, with a little thought..
    
    This one, 4_vs_6 is good for exercising wireless hw queues,
    and, as it turns out, useful for the shaper, too.
  3. Removed too much of the still useless ingress code

    Dave Taht committed Apr 15, 2012
    I'm just not convinced ingress is the right thing.