Just on the off chance anyone recognises the symptoms (yeah I know long shot but you never know ;-) ) ...
I have several java processes running against a clustered Oracle database, the SQL is in some cases trivial. On certain days I get problems where it appears all the SQL takes ages to run >50s for normally sub second queries and it looks like the different SQL acting on different tables returns at the same instance even though executed at different start times but within a small window. The DBA believes the SQL in question is executing incredibly fast but stuck waiting for a commit(s) from the network, the connection is auto commit. The database is showing no signs of strain (very under worked), the network guy says the network is under light load (we're watching network traffic for issues) and the infrastructure bod says disk latency is low. The driver I am using is the one via OCI client ie not the thin driver. Some days we get only get small acceptable problems when the database has its backup task executing, other days melt down.
Any thoughts, wild guesses ...
"Eagles may soar but weasels don't get sucked into jet engines" SCJP 1.6, SCWCD 1.4, SCJD 1.5,SCBCD 5
Since no one commented ;-) I'll post the solution in case anyone else finds the same problem. The issue was the redo logs that are written to on a commit were set to switch way too frequently. So although the database, app and network all reported extremely fast performance everything was blocked on these logs.
I’ve looked at a lot of different solutions, and in my humble opinion Aspose is the way to go. Here’s the link: http://aspose.com