VMM - is the virtual memory manager, a process that is part of the kernel
and is responsible for managing
memory.

Virtual memory is real memory plus disk space (both addresses to data on
disks and paging space)

Non-Computational memory pages are pages of real memory that contain data
that has a persistent location
on disk (executable code of a command, data from a file, etc. not from
data stored in paging space).

Computational memory pages are pages of real memory that do not have a
persistent location on disk, they
are pages of memory needed by a process when it is started and no longer
available when the process
terminates. These are pages in memory that contain things like process
stack and heap, and malloced
space. When memory becomes over committed these pages need to be stored
someplace temporarily
and that is on paging space.
tb steps
 
vmstat -I 2 10                                                         
vmo -a                                                                 
lsps -a                                                                 
svmon -G         

is your swapspace low if so this may help. Defect IY62224 adds a new LRU tunable, lru_file_repage.  This is
on by default, which maintains the current behavior.  The new behavior     
when it is off is that file vs computational repage counts are ignored 
in determining whether to page out only file pages.                     
                                                                       
vmo -p -o lru_file_repage=0  (NO REBOOT NEEDED)                         
This document explains more about lru_poll_interval


I recommend disabling lru_file_repage=0                                 
and sentting lru_poll_interval=10                                       

#!/usr/bin/ksh
# Robert BInkley
#Collect vmstat and ps to detect memory leak
# Version History
# 07/24/02: Split perf data into separate files
#LOGDIR=/tmp/rob ## Change to use in your enviorment
#LOGDIR=/system/perf/robi
if [ ! -d $LOGDIR ]
then
mkdir -p $LOGDIR
fi
HN=`hostname`
DATE=`date +"%y.%m.%d.%H.%M"`
ps -ef> $LOGDIR/$HN.$DATE.ps
while true
do
(     
date
vmstat 5 3
echo
#ps vg
#echo
lsps -a
echo
iostat -d 5 2|egrep "hdisk0|power"
echo
) >>$LOGDIR/$HN.$DATE.perf 2>&1

#SVMON

(
date
/usr/bin/svmon

/usr/bin/svmon -Pt 5|grep -ip "pid"
echo
) >> $LOGDIR/$HN.$DATE.svmon 2>&1

#FILEMON

date >>$LOGDIR/$HN.$DATE.filemon
(
date
filemon -Opv,lv
sleep 240
trcstop
echo
) > $LOGDIR/$HN.filemon.tmp 2>&1

#VMTUNE
(
date
/usr/samples/kernel/vmtune
echo
) >> $LOGDIR/$HN.$DATE.vmtune 2>&1
sleep 10
grep -ip "Most Active" $LOGDIR/$HN.filemon.tmp
>>$LOGDIR/$HN.$DATE.filemon
echo >>$LOGDIR/$HN.$DATE.filemon
done &

Well, one quick and dirty method would be:

ps -e -o pid,vsz,args

This would give a formatted output of the PID, the virtual size of the process in kbytes, and the command line. The memory calculations that the ps command use aren't all that accurate. But it should work to identify a growing process.

Another command is:

ps gv

Look for the RSS column. This is the resident set size (real memory usage) of the process in kbytes.

Still, the svmon command gives the most accurate picture of memory use.