mailboxd_java_heap_memory_percent for 32GB ram server?

[gettyless]

We've noticed that top reports back a steady load of 3 since we migrated about 50+ accounts over from postfix to this new zimbra server. It's been working like a champ, I just want to tidy up things if possible or maybe just optimize/lower the load/memory/cpu usage for Mother Earth.

My questions are:

1) why the constant load of 3?

2) since I have 32GB, maybe lowering mailboxd_java_heap_memory_percent to say 20 or 15 percent would make sense? The last server I worked on, had lesser hardware and 4x the number of users.

3) memory usage is constantly about 90% -- IIRC this is "by design" but I forget the correct reasons. Something to do with how java works and how it needs to cache a lot of memory to speed up java'ish stuff.

4) we have heavy IMAP usage here (only one outlook user to my relief), is there anything I should consider tweaking? Nothing is broke so for this question, I'm mostly looking for what I can think over, not really take action on.

Specs of the box are:

32GB memory
RAID 10
kernel 2.6.24-23
zcs-NETWORK-5.0.15_GA_2851.UBUNTU8_64.20090310194234

Usage observed at 2pm on a weekday last week was 81 IMAP, 10 web.

Here's some of the settings I have, all of which are default from zimbra:

data_source_max_message_memory_size = 2097152
mailboxd_java_heap_memory_percent = 30
mysql_memory_percent = 40
javamail_imap_debug = false
javamail_imap_enable_starttls = true
javamail_imap_timeout = 60
javamail_pop3_debug = false
javamail_pop3_enable_starttls = true
javamail_pop3_timeout = 60
javamail_smtp_debug = false
javamail_smtp_enable_starttls = true
javamail_smtp_timeout = 60
mailboxd_java_heap_memory_percent = 30
mailboxd_java_home = ${zimbra_java_home}
mailboxd_java_options = -client -XX:NewRatio=2 -Djava.awt.headless=true -XX:MaxPermSize=128m -XX:SoftRefLRUPolicyMSPerMB=1
mailboxd_truststore = ${zimbra_java_home}/lib/security/cacerts
zimbra_java_home = ${zimbra_home}/${zimbra_java_path}
zimbra_java_path = java

[gettyless]

also here's my innodb stats:

----------------------
BUFFER POOL AND MEMORY
----------------------
Total memory allocated 8463462130; in additional pool allocated 1048576
Buffer pool size   464192
Free buffers       397272
Database pages     63535
Modified db pages  0
Pending reads 0
Pending writes: LRU 0, flush list 0, single page 0
Pages read 51518, created 12040, written 3238310
0.00 reads/s, 0.00 creates/s, 18.44 writes/s
Buffer pool hit rate 1000 / 1000

[gettyless]

I guess this is some random java memory usage/leak issue? I upgraded to zcs-NETWORK-5.0.16_GA_2921.UBUNTU8_64.20090429025807 on July 3rd and since then load is way down to under 1 even during peak hours (at least for today, the first normal business day afterwards).

No other settings were changed. I probably should have just rebooted before I upgraded to verify behavior.

hmmm.