Multiple freelists ... do they really reduce contention

Here's the situation: My application has multiple clients, which can all be performing the same operation at the same time. In stress testing, as expected, the response time degrades (although overall throughput does increase). When I look at a graph of completion times, I see that multiple requests complete at the same second -- indication to me that one is holding a lock, and the others complete when that lock is released.

Reading about the FREELISTS parameter of the STORAGE clause, it seemed to me that my problem was blocking for free blocks, and that I could solve this problem by creating N freelists on the affected tables/indexes, where N = the number of simultaneous clients.

I did this, but nothing changed: execution times were very similar (any difference attributable to test environment), and the graph still showed this blocking behavior.

So, I'm wondering if I have the correct interpretation of FREELISTS, if anyone else has used them and seen an improvement in concurrent access, and any other places I should look to determine where the locks are occuring.

Thx,
-kdg Received on Fri Feb 06 1998 - 00:00:00 CST