Create a tool for USL analysis

[GoogleCodeExporter]

Create a tool that, given a file of load versus throughput, will do Universal 
Scalability Law analysis on it.  It should assume the existence of gnuplot on 
the system.

Original issue reported on code.google.com by baron.schwartz on 15 Oct 2010 at 11:24

[GoogleCodeExporter]

This issue was updated by revision r242.

Initial scratch file

Original comment by baron.schwartz on 15 Oct 2010 at 11:25

[GoogleCodeExporter]

This issue was updated by revision r244.

Initial proof of concept

Original comment by baron.schwartz on 15 Oct 2010 at 1:36

[GoogleCodeExporter]

This issue was updated by revision r245.

Add a sample file

Original comment by baron.schwartz on 15 Oct 2010 at 1:39

[GoogleCodeExporter]

This issue was updated by revision r246.

Add another sample, this one from SHOW GLOBAL STATUS

Original comment by baron.schwartz on 15 Oct 2010 at 1:50

[GoogleCodeExporter]

This issue was updated by revision r247.

Add test suite and sample files for usl

Original comment by baron.schwartz on 15 Oct 2010 at 2:36

[GoogleCodeExporter]

This issue was updated by revision r248.

Pretty solid working implementation

Original comment by baron.schwartz on 15 Oct 2010 at 8:20

[GoogleCodeExporter]

This issue was updated by revision r249.

Change X(concurrency) to C(concurrency/capacity)

Original comment by baron.schwartz on 16 Oct 2010 at 1:13

[GoogleCodeExporter]

This issue was updated by revision r250.

Add the -p <prefix> option

Original comment by baron.schwartz on 16 Oct 2010 at 1:16

[GoogleCodeExporter]

This issue was updated by revision r251.

Add support for EPS files with -t, and improve the graphs a lot

Original comment by baron.schwartz on 16 Oct 2010 at 8:57

[GoogleCodeExporter]

This issue was updated by revision r252.

Add -t pdf for convenience

Original comment by baron.schwartz on 17 Oct 2010 at 3:34

[GoogleCodeExporter]

This issue was updated by revision r255.

Add documentation

Original comment by baron.schwartz on 18 Oct 2010 at 7:34

[GoogleCodeExporter]

This issue was updated by revision r256.

tweak the usage info

Original comment by baron.schwartz on 18 Oct 2010 at 7:34

[GoogleCodeExporter]

This issue was updated by revision r257.

After quadratic fitting with transformed data, re-fit against the USL directly 
for the best fit.

Original comment by baron.schwartz on 19 Oct 2010 at 3:03

[GoogleCodeExporter]

This issue was updated by revision r258.

Update the usl docs

Original comment by baron.schwartz on 19 Oct 2010 at 3:17

[GoogleCodeExporter]

TODO: need to make it optionally make logarithmic X axis graph.  Also, apply -i 
and -n and so on in the main body, not in the converter itself, so that I can 
convert a file and then manipulate it separately.

Original comment by baron.schwartz on 20 Oct 2010 at 3:30

[GoogleCodeExporter]

This issue was updated by revision r259.

Add -l option

Original comment by baron.schwartz on 20 Oct 2010 at 2:13

[GoogleCodeExporter]

This issue was updated by revision r260.

Update docs for -l

Original comment by baron.schwartz on 20 Oct 2010 at 2:15

[GoogleCodeExporter]

This issue was updated by revision r261.

tweak globalstatus scripts in docs

Original comment by baron.schwartz on 20 Oct 2010 at 2:49

[GoogleCodeExporter]

TODO: I want it to entirely skip samples that have too-large concurrency, as 
defined by -m.  I don't want it to just average the concurrency across the 
other samples; it will introduce skew into the throughput for that sample, too.

Original comment by baron.schwartz on 20 Oct 2010 at 2:51

[GoogleCodeExporter]

This issue was updated by revision r263.

add the -a and -e options

Original comment by baron.schwartz on 21 Oct 2010 at 1:24

[GoogleCodeExporter]

This issue was updated by revision r264.

Add the -a and -e options to the documentation

Original comment by baron.schwartz on 21 Oct 2010 at 1:26

[GoogleCodeExporter]

This issue was updated by revision r265.

minor typo in docs

Original comment by baron.schwartz on 21 Oct 2010 at 3:23

[GoogleCodeExporter]

This issue was updated by revision r267.

Add a few samples of TCP dumped traffic, and add a new -c converter for 
tcpdump, and tests for it

Original comment by baron.schwartz on 24 Oct 2010 at 9:06

[GoogleCodeExporter]

This issue was updated by revision r268.

Add a usl-prefix to all filenames

Original comment by baron.schwartz on 24 Oct 2010 at 11:48

[GoogleCodeExporter]

This issue was updated by revision r269.

docs for the tcpdump converter

Original comment by baron.schwartz on 24 Oct 2010 at 11:57

[GoogleCodeExporter]

This issue was updated by revision r270.

docs for the -P option

Original comment by baron.schwartz on 24 Oct 2010 at 11:58

[GoogleCodeExporter]

This issue was updated by revision r271.

Fix another filename prefix with usl-.
Fix a bug in -c tcpdump when the client sends multiple packets before the 
server responds.

Original comment by baron.schwartz on 25 Oct 2010 at 1:23

[GoogleCodeExporter]

This issue was updated by revision r274.

Add the -o parameter

Original comment by baron.schwartz on 30 Oct 2010 at 4:00

[GoogleCodeExporter]

This issue was updated by revision r275.

Add the -o parameter docs

Original comment by baron.schwartz on 30 Oct 2010 at 4:01

[GoogleCodeExporter]

This issue was updated by revision r276.

Add the -r, -L, and -T options for controlling how gnuplot renders graphs

Original comment by baron.schwartz on 30 Oct 2010 at 11:42

[GoogleCodeExporter]

This issue was updated by revision r286.

Add -R option

Original comment by baron.schwartz on 11 Nov 2010 at 3:34

[GoogleCodeExporter]

This issue was updated by revision r289.

Fix spaces-to-tabs brokenness in a HERE-doc

Original comment by baron.schwartz on 8 Dec 2010 at 11:12

[GoogleCodeExporter]

This issue was updated by revision r290.

Add -C parameter to usl so regression can be used to find C(1)

Original comment by baron.schwartz on 18 Dec 2010 at 12:38

[GoogleCodeExporter]

This issue was updated by revision r291.

Remove the option, and always do a regression for C(1)

Original comment by baron.schwartz on 18 Dec 2010 at 12:43

[GoogleCodeExporter]

This issue was updated by revision r292.

Remove the option, and always do a regression for C(1)

Original comment by baron.schwartz on 18 Dec 2010 at 12:43

[GoogleCodeExporter]

This issue was closed by revision r293.

Original comment by baron.schwartz on 30 Jan 2011 at 11:30

• Changed state: Fixed

[GoogleCodeExporter]

This issue was updated by revision r294.

docs for -X

Original comment by baron.schwartz on 30 Jan 2011 at 11:31