[dtaht/deBloat] : Basic 4Mbit testing and partial log results

GitHub noreply at github.com
Thu Jan 12 22:10:57 PST 2012


  Branch: refs/heads/master
  Home:   https://github.com/dtaht/deBloat
  Commit: d0c4d4f8b7f560962a3541760876cab48bbea0e4
      https://github.com/dtaht/deBloat/commit/d0c4d4f8b7f560962a3541760876cab48bbea0e4
  Author: Dave Taht <d at taht.net>
  Date:   2012-01-12 (Thu, 12 Jan 2012)

  Changed paths:
    M src/debloat
  A test/rtt/itest
  A test/rtt/notes.org

  Log Message:
  -----------
  Basic 4Mbit testing and partial log results

I put up some test results in the test/rtt directory. I
have more detailed logs too, but they point to SFQ working
really well to preserve interactive streams.

Trying to figure out to what extent red is working vs sfq
doing it's job is hard. I need to be monitoring end-to-end
stuff, and analyzing the packet captures on both sides.

My primary concern is with byte oriented red here and watching
iperf misbehave, taking 67 seconds to conclude a test that I
had specified 60 for on the 10 stream test. Did I end up with
~7 seconds of buffering in the high rate streams, delaying the
FIN/ACK process?

Or is it just iperf misbehaving? netperf seems not to...





More information about the Cerowrt-commits mailing list