It's a bit of a hybrid system. OLTP throughout the day, as well as having large amounts of data bulk-loaded throughout the day, after which that bulk-loaded data is read-only. One of the reasons I started down this path was when I saw ad-hoc queries from some marketing/sales guy firing off parallel queries. That's when I though Doug's approach sounded pretty good. Cut off all default parallel usage, and those applications that want to use parallel can supply a query hint. However I can't say that the parallel queries were really causing any performance problem. Of course, my current parallel_max_servers is 8, and the formula would indicate that it should be 80. Don. On 5/15/07, Dennis Williams <oracledba.williams@xxxxxxxxx> wrote:
Don, The other thing you haven't mentioned is what your database workload is like. For example, if it is an OLTP system, you don't want a parallel process to max out the machine, but force all your other users to wait. On the other hand, for a DSS system, that may work just fine - the big query may be the one the boss needs and usage is pretty sporadic anyway. Dennis Williams
-- Don Seiler oracle blog: http://ora.seiler.us ultimate: http://www.mufc.us -- //www.freelists.org/webpage/oracle-l