Peter, >>"so I don't think there is much user wait time in here." The tracefile clearly shows that of the 600 seconds covered by it (the tracefile), the majority - 391 seconds - was spent at the client-side (ie. not in the dbms, but user wait/think time). >>"I am beginning to think that its the shear volume of calls thats the real killer here." The shear volume of calls is indeed the killer for the remaining 209 seconds... (this, btw, is also a problem in a typical Java-centric architected application: the number of calls from the middle tier to the database tier is killing perfomance). Toon -- //www.freelists.org/webpage/oracle-l