I would not be so quick to modify the listener.

In my experience there are two other areas that have a higher probability of being the problem.

The first is the network. Make sure the users are not seeing a sort of venturi effect where their 100MB NICs are on a 10MB line or switch port. This can cause packet drops which appear as timeouts.

The second is to make sure you are not approaching a system limit on the number of connections/sessions/processes. This is a performance limit, not a parameter one. I have one system that is supposed to handle 200+ connections (enough memory, etc) but bogged down miserably at about 180. This can also cause timeout problems.


You did not say what your OS is. If it is one of the Unixes, what is the box stress level while this is happening?