>If you are on a platform with a decent thread library the worker mpm
>is a good bet. mod_dav + svn is thread safe.
Is Debian Linux with kernel 2.4.21 considered to have a good thread library?
>It kind of depends on how much memory you have, what your usage pattern
>looks like etc. Try switching to the worker mpm first and then decide
>how much memory you can live with being (permanently) taken up by httpd.
>Devide that number by the average number of httpd processes (with worker
>default config and no heavy load: 2).
Okay, thanks. Is there any way to know when allocating more memory to
httpd would speed things up? In other words, how do I tune the number,
besides just deciding a priori that httpd should get 100mb or something?
I'm actually trying to figure out why svn seems slow (slow is defined as
relatively high latency and low throughput...each command takes a while to
run and doesn't move data very fast across the network), and I thought I'd
start with this. However, according to atsar the machine's not paging at
all, so I doubt the memory is the problem (but it's good to know how to
constrain the beast, so the above is valuable :). The httpd process is
only taking <10% of CPU when an svn command is running as well, and the net
isn't remotely taxed. So, I guess I need to look into disk IO, but I don't
see why caching wouldn't take care of that (the machine was running with
200mb devoted to cache according to the last top I posted). I wonder if
running over https is slowing things down because of the SSL negotiation on
each connection? But, it seems like keepalive should reduce that. I'm
totally inexperienced at profiling things on unix, obviously. :)
Thanks a lot,
Chris
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Tue Jul 1 09:20:01 2003