[dtaht/deBloat] d3a645: Pure QFQ AQM solution for wireless and 100Mbit eth...

noreply at github.com noreply at github.com
Tue Dec 27 04:15:06 PST 2011


  Branch: refs/heads/master
  Home:   https://github.com/dtaht/deBloat

  Commit: d3a6458aa659860f48e83ab61693de46d373df45
      https://github.com/dtaht/deBloat/commit/d3a6458aa659860f48e83ab61693de46d373df45
  Author: Dave Taht <d at taht.net>
  Date:   2011-12-27 (Tue, 27 Dec 2011)

  Changed paths:
    A src/staqfq.lua
  A src/tcparse.lua

  Log Message:
  -----------
  Pure QFQ AQM solution for wireless and 100Mbit ethernet

The staqfq.lua code can now be renamed and made standard
in /etc/network/if-up-* on a normal desktop or server box
running a bql enabled kernel and git head of iproute2.

In addition to being the first version of this code to
actually work well (reducing interstream latencies by
two orders of magnitude on 100Mbit ethernet on a bql
enabled box)

This is the first version of this code re-written in lua.
Lua's advantages:

* More powerful than either C or shell (closures! yea!)
* Std language on openwrt, easily used on other systems
* Tiny language in either case

The code is now far more easy to read than shell and nearly
as fast as C. It also now uses tc's bulk load facility, and
loads the rules in under a second.

Also included in this commit is a tool to make reading tc -s qdisc
output mildly easier.


  Commit: 1665ae746e81db113c25d20d9a60be91e59930dc
      https://github.com/dtaht/deBloat/commit/1665ae746e81db113c25d20d9a60be91e59930dc
  Author: Dave Taht <d at taht.net>
  Date:   2011-12-27 (Tue, 27 Dec 2011)

  Changed paths:
    A src/cap.lua
  A src/renumber.lua

  Log Message:
  -----------
  Network renumbering and (one day) interstream analysis scripts

While I'm on this lua kick, I'm also prototyping several other
lua based scripts and building up infrastrructure.

cap.lua will hopefully one day read capture files and produce
a interflow analysis graph.

renumber.lua is a start towards making network renumbering
simpler (with a gui) on cerowrt.


  Commit: dff94deb9368b075205ccc684f98a612ba1e8ab2
      https://github.com/dtaht/deBloat/commit/dff94deb9368b075205ccc684f98a612ba1e8ab2
  Author: Dave Taht <d at taht.net>
  Date:   2011-12-27 (Tue, 27 Dec 2011)

  Changed paths:
    A models/htb-bt-qfq-red-4mbit.lua
  A models/htb-bt-qfq-red-desktop
  A models/qfq-red-desktop
  A openwrt/blacklist
  A openwrt/classify/diffserv
  A openwrt/functions.sh
  A openwrt/stratagies.db
  A openwrt/strategy
  A samples/matches
  A samples/p2phammer
  A samples/tbfversion
  A samples/tcpgraph.cgi
  A samples/tcpgraph.css
  R src/Makefile
  A src/NOTES.org
  R src/htb-bt-qfq-red-4mbit.c
  R src/tcbuilder.c
  A test/lab
  A test/lab2
  A wip/batch-htb-bt-qfq
  A wip/batchbtqfq
  A wip/btqfq
  A wip/eqfq.gso
  A wip/eqfq.red.lfn
  A wip/eqfq.simplewireless
  R wip/tbfversion
  A wip/tc_index
  A wip/tcpparsestat.pl
  A wip/wshaper.htb

  Log Message:
  -----------
  Brain dump from my test machine

This a bunch of wip, experiments, ideas, etc, that need
to live in one repo, on multiple machines so I can clean
it up.

Also nuked the c version of the shaper ideas, lua is superior.


  Commit: 3e62ddece358c0b5f7d202b63a75b64acb9825e6
      https://github.com/dtaht/deBloat/commit/3e62ddece358c0b5f7d202b63a75b64acb9825e6
  Author: Dave Taht <d at taht.net>
  Date:   2011-12-27 (Tue, 27 Dec 2011)

  Changed paths:
    A test/README.org
  A test/bench
  A test/itest
  A test/itest2
  A test/itest4
  A test/itestc
  A test/runtest
  A test/stats.org

  Log Message:
  -----------
  Slammed a bunch of adhoc tests I run into the repository

As I've finally settled on a set of useful benchmarks for measuring
latency under load, it's time to clean them up, and make them
repeatable for others.

That does not mean that that's what is in this commit.

I'm not proud of committing crappy code, but when you are crashing
your main development box on an hourly basis, it is a rather
good idea to NOT keep anything in just one place.


  Commit: 6cf942e4193e28527844e3bc84e5cbf75034e74e
      https://github.com/dtaht/deBloat/commit/6cf942e4193e28527844e3bc84e5cbf75034e74e
  Author: Dave Taht <d at taht.net>
  Date:   2011-12-27 (Tue, 27 Dec 2011)

  Changed paths:
    M Makefile
  A src/Makefile
  M src/staqfq.lua

  Log Message:
  -----------
  Noted that the multicast shaping feature wasn't actually working

I would like very much to handle multicast right, but apparently
my mental model of the endianess of the multicast mac address is
backwards, upside down, or wrong.


Compare: https://github.com/dtaht/deBloat/compare/2c8c2ac...6cf942e


More information about the Cerowrt-commits mailing list