Re: [tlug] Apache MPM performance

["Nguyen Vu Hung" <vuhung16plus+shape@example.com>]

2008/3/11, Josh Glover <jmglov@example.com>:
> On 11/03/2008, Nguyen Vu Hung <vuhung16plus+shape@example.com> wrote:
> What is the difference between valgrind massif and regular old valgrind?
By default, valgrind runs memchek and it is equivalent to

valgrind --tool=memcheck

which will check memory leak.

if we run

valgrind --tool=massif[1]

it will profile the heap usage of the program to be run.

I use massif as a "unit test" checking how much, and how long the
memory is used.

But when we do an "intergration test", we have to another method.

>
>  valgrind and vmstat are really for two different things; the former is
>  for optimising and debugging your code, the latter for watching the
>  effect of your code plus the full environment around it (i.e. Linux
>  kernel + Apache + mod_perl + your code).

Actually, valgrind is not for debugging but profiling and optimization.

My code uses C/C++ and PHP, and I can massif it by running

valgrind --tool=massif /path/to/file.php

from command line.
>
>  Solaris *must* have a tool like vmstat, as all vmstat does is
>  aggregate data available through the /proc filesystem, which Solaris
>  and all good Unices support. Anyone know what it is called and how to
>  use it?
It is vmstat same as Linux vmstat.

[1] http://valgrind.org/docs/manual/ms-manual.html

-- 
Best Regards,
Nguyen Hung Vu ( Nguyễn Vũ Hưng )
vuhung16plus{remove}@example.com
An inquisitive look at Harajuku
http://www.flickr.com/photos/vuhung/sets/72157600109218238/