Details:
Hibernate 2.1.4
Oracle 8.1.7
Oracle JDBC classes12.zip
Java JDK 1.3.1

I am trying to stress test Hibernate. I have an existing application that
gets log records and inserts them into an Oracle database. This is a Corba application that receives insert requests from a listener app and sends the insert to Oracle. The current production application uses an Oracle data source
, connection pool and jdbc to insert the data. I am testing Hibernate
and doing the same processing except with the c3p0 connection pool. To keep the test the same I am using the jdbc connection rather then trying to persist a java object.

The production version of the code can sustain up to 330 inserts per second.
The hibernate version of the code maxes out at around 180 tps.

Both versions of the test run on the same hardware/dbms combination and use the same process.

cs = getClientSession(); // Calls SessionFactory.openSession()
Connection con = cs.connection();
Statement stmt = con.createStatement();
stmt.executeUpdate(sql);

stmt.close();
cs.close();



My basic question is are there overheads in Hibernate that I need to be aware of that can be optimized or turned off?

Enabling batching might help. Set it to a non zero value.

hibernate.jdbc.batch_size

try what lneelaka said,
also show c3p0 config

If you are running inside an appserver, use the appserver connection pool.

ie. don't test c3p0 vs. appserver connection pool, they have differenct behaviors.

Oh, also, I don't see you commit the txn! Note that Hibernate returns connections that have a txn started!