Re: JDBC Driver contention

Hi. I have observed the same exact behavior with my OCI C program, which performs batch INSERT's. It simply doesn't scale well at all. In my case, peak performance is obtained with 2 worker threads under Solaris and 1 worker thread under Linux. Launching multiple instances of my program concurrently provides much better performance on both platforms (near-linear scaling), proving that the root problem has nothing to do with the database, CPU bottlenecks, I/O bottlenecks, etc. I've also added timing code, the results of which suggest that the problem is isolated to my OCIStmtExecute() call. For this section of code, increasing the number of threads beyond 2 results in increased CPU user time (expected) AND increased real time (unexpected). Does anyone have a clue?

--
Posted via http://dbforums.com