Our product uses distributed computing to share load among various
servers.
The main server receices requests for client and dispatches them to
the worker server for doing the actual work.

The main server does connection pooling with the worker server to
avoid recreating the connection.

We see very high kernel time when we reuse the same connection.

However, if we close the connection after every request to the worker
server and that forces it ot open new connection for every request
than the main server show very low kernel time.

This kernel time avoids us from pushing the main server to receive
more request even when the total cpu usage is only 60%


Any tool we would use to figure out the reason for the high kernel
time.

Do we need to set any socket option.

Any Suggestins why that would happen.

It happens on HP-UX 11.00 and 11i