Server load and MYSQL

[Aaron]

Hi all.  I'm looking for a second opinion. (or a third, fourth, fifth....)

Here's the situation:  I'm running a server with the following specs:  Dual P3 1.3Ghz, 1GB ECC 
Registered RAM, Dual 18GB SCSI Ultra 160 10K RPM drives, Adaptec 29160 (the 64bit version)...  <-- 
Important specs. :)

This server is running RedHat 7.3 (with an upgraded kernel. 2.4.18-17smp) and all the normal stuff. 
(apache, bind, mysql, php, etc etc etc) The web site it serves gets 20,000 unique visitors a day 
and the databases are roughly 6GB.

For the last two months I've had no end to the problems.  First, the machine would crash randomly, 
once a day.  I figured out that it was a Kernel bug, hence the reason for the upgrade.

Second, every once in a while (once every 3 or 4 days) Apache would crash and I'l look at the load 
to find it at 300 - 400.  It wouldn't let me soft reboot it so I'd have to shut it down.

I moved the databases and the domlogs to the second drive and that seems to have relieved some of 
the load and stopped the /var and /usr partitions from hitting 100% full every day, but it's still 
not uncommon for me to see loads in the 8-9 range.  I'd like to get the machine a little more 
stable.

So my questions....  

1.  Does anyone have an opinion on switching the journaling system to writeback?  It looks like 
that's better for database writes but not so good for data integrity.

2.  I saw something about MySQL taking 28MB of memory for each process it spawns and that that 
could be reduced and would lighten the load, but I can't get any details.  Also, in my my.cnf file 
it says it's set to 1MB.  Is there something I'm overlooking?

Any help, suggestions, opinions, etc would be greatly appreciated.

Aaron