Re: Apache::Session::MySQL lock troubles

[Dan Axtell <da...@snet.net>]

I've been using Apache::Session::MySQL for a while, but I've been trying to 
lower the amount of database I/O in some applications that experience spikes 
in server traffic.  So I came up with the idea of using Apache::Session::File 
with the files on /dev/shm, with a cron job to clear up old or expired 
sessions.  Preliminary benchmarks found this to be about 7-8 times faster 
than MySQL.

I'm curious if anyone else has tried this, and if they can think of any 
reasons why this would be a bad idea.

Dan

Re: Apache::Session::MySQL lock troubles

[Perrin Harkins <pe...@elem.com>]

On Thu, 2006-03-02 at 14:10 -0500, Dan Axtell wrote:
> I've been using Apache::Session::MySQL for a while, but I've been trying to 
> lower the amount of database I/O in some applications that experience spikes 
> in server traffic.  So I came up with the idea of using Apache::Session::File 
> with the files on /dev/shm, with a cron job to clear up old or expired 
> sessions.  Preliminary benchmarks found this to be about 7-8 times faster 
> than MySQL.

Sure, it's local.  Using BerkeleyDB would also be faster than MySQL.
The downside of your approach is that it can't scale across a cluster of
machines, and has no on-disk backup.  If you only have one machine, it's
probably fine.

- Perrin

Re: Apache::Session::MySQL lock troubles

[Jonathan Vanasco <jo...@2xlp.com>]

you could also try this:

	store to /whatever/
	read from memcached
		failover to /whatever/
	
assuming you have the memory for it, it should handle spikes very well