Michael Peters wrote:
prefork will use a different process for each Apache child and thus a
different Perl interpreter for each process. Normally this would mean
a lot of memory usage, but since Linux uses COW (copy-on-write) memory
as long as you preload you'll actually see about the same memory usage
as with worker/threads.
By "preload" you mean loading all perl modules I'm using in scripts?
Yes, I do preload all the modules at startup.
And if you have a multi-core or multi-processor system, having
sepeparate processes that know nothing about each other should result
in less contention (they aren't sharing internal Perl structures).
Plus you don't have to worry about trying to make sure every module
you use from CPAN is thread-safe.
Aahhh, thanks for this hint: "thread-safe modules". Now I'm sure I'll
use prefork.
Thanks Michael.
P.S. not mod_perl related but .. For a few weeks I thought about
switching to Ruby, Ruby on Rails after 5 years of mod_perl... but still
decided to use mod_perl in the future, that's why I'm very cautious
about mod_perl and want to be sure it will run smoothly.
