Portal not stable - goes down during normal config operation

We have a case of Portal system frequently going down during normal

configuration activities:

- activities in Theme Editor

- changing and saving Iview

I looked through the logs but I struggled to find errors which would

explain the cause of this problem

I wanted to implemented recommanded parameters form note 716927(see below) but I am not 100% where I should enter them.

When I go to Config Tool should they be enetered on:

- Instance level

- or Server Level

Any feedback would be greatly appreciated

-Xms2048M

initial heap size

-Xmx2048M

maximum heap size

-Xconcurrentbackground<n>

controls how many background helper threads participate in GC. Must be equal to the number of processors in an SMP or LPAR.The default value is 1.

-Xloratio0.3

controls how much of the total heap ( -Xmx setting ) is reserved for large objects. Default is 0.05, reserving 5% .

-Xparroot

enables early, parallel scanning of roots to shorten GC pause times

-Xconcurrentlevel6

controls the timing of starting concurrent GC work

-XSoftRefThreshold0

controls how many GC cycles objects kept alive by softreferences survive

-Xconmeter2

instructive parameter for the concurrent GC accurate triggering and metering

-Xk17000

the maximal number of classes the kCluster will contain, default is 1280 (each of 256 bytes). kCluster - 1st object of the heap

-Xp2048K,128K

the initial size of the pCluster and the overflow size. pCluster - 2nd object of the heap

-verbose:Xgccon

for additional mostly concurrent data