Initial benchmarks for checkout performance improvements

Matthew D. Fuller fullermd at over-yonder.net
Thu May 31 22:03:13 BST 2007


On Thu, May 31, 2007 at 04:50:06PM -0400 I heard the voice of
Aaron Bentley, and lo! it spake thus:

Some more ministat runs, for fun.  Note that all of these showed no
difference at 95% confidence.  Whee   :)


> Performance of the top-rename strategy was similar to copytree:

% ./ministat -sw73 copytree top-rename 
x copytree
+ top-rename
+-------------------------------------------------------------------------+
|x             +  +           +       x           +++  +x+x* +xx  x x     |
|                                |___________________A_____M_____________||
|                           |________________A______M_________|           |
+-------------------------------------------------------------------------+
    N           Min           Max        Median           Avg        Stddev
x  10           4.7         11.38         10.51         9.923      2.005875
+  10          6.14         10.76         9.785         9.109     1.7293958
No difference proven at 95.0% confidence



> In this run of benchmarks, cp was actually slower than copytree or
> top-rename.

% ./ministat -sw73 cp copytree
x cp
+ copytree
+-------------------------------------------------------------------------+
|                                            x x          +               |
|+                                    +   xxxx x  x     * ++  ++  + +  x  |
|                                       |_____M__A_______|                |
|                                |___________________A_____M_____________||
+-------------------------------------------------------------------------+
    N           Min           Max        Median           Avg        Stddev
x  10          8.78         11.68          9.19         9.493     0.8760397
+  10           4.7         11.38         10.51         9.923      2.005875
No difference proven at 95.0% confidence



> I also implemented a second strategy: "rename-multi".  [...] This does not
> seem to be faster than "top-rename".

% ./ministat -sw73 top-rename rename-multi 
x top-rename
+ rename-multi
+-------------------------------------------------------------------------+
|                                                      +                  |
|x   x+            +   +x                             x+x x + x++x  x++ x |
|                   |_________________________A__________M_______________||
|                        |______________________A________M______________| |
+-------------------------------------------------------------------------+
    N           Min           Max        Median           Avg        Stddev
x  10          6.14         10.76         9.785         9.109     1.7293958
+  10          6.47         10.67          9.81         9.226     1.5182168
No difference proven at 95.0% confidence



-- 
Matthew Fuller     (MF4839)   |  fullermd at over-yonder.net
Systems/Network Administrator |  http://www.over-yonder.net/~fullermd/
           On the Internet, nobody can hear you scream.



More information about the bazaar mailing list