Oracle Parallel Server configuration (40 instances on a single database ???!!!?) will help you *absolutely nothing* in speeding your long-running queries! The query will still be processed by a single instance (on a single machine). OPS is ment for higher availability and/or scalability of the system, not for speeding up queries.

In your case you must first determine what is your bottleneck: CPU, I/O, memory, network, ....., or simply inefficiently written queries. After determining that, you can concentrate on the offending part of the sistem and improve it accordingly.