merge command, pq slaves and sorting

I have a huge merge command that does a hash join, 50 window sort steps and an associated FTS.
With even a degree of 2 on the involved tables, it spawns 20 PQ slaves and never completes.
With noparallel the query completes in 3-4 minutes.

Can someone please verify...If the FTS spawns a number of slaves based on the degree, does each sort operation also get that number of slaves, is that correct?

parallel_max_servers = 20 and thus why only 20 are used I belive, but I think the job would actualy want about a 100 slaves.

Hmmm, thanks

-- 
..
David


--
http://www.freelists.org/webpage/oracle-l