JDBC: Connection object's Autocommit value for SELECT

Hello,
please give feedback on the following issue I've seen.

We have written a java application that is being deployed against websphere. We use the websphere connection pool mechanism, which means that we never actually disconnect from the database (unless websphere decides to do so because of a timeout or so).
The application connects to Oracle using the 8i Oracle JDBC driver.

New Oracle connection objects are in autocommit mode which means that, unless we set autocommit to false (which we do of course before issueueing the first DML statement), the database still has to process a lot of commits even when our webapplication does mainly SELECTs.

What I've seen is that, under a heavy load, the most waited for event is "control file parallel write" (besides "cache buffer chains" but that's a different story). I suspect the many commits caused by the JDBC driver are due to this. Anybody seen this before? Why not always change the default setting for autocommit to false when running against Oracle databases and make sure commits are being done explicitely after DML statements?

Thanks for your feedback,

Andre van Winssen Received on Mon Apr 15 2002 - 10:17:17 CDT