we have master tables as large as 8GB and we only have 500MB in the buffer
cache, plus other processes going on. I think the biggest slow down comes that
while we are doing the large ingestions those blocks are getting pushed out of
the buffer cache and being re-queried by the anti-join.

I ran statspack, but Im just digging into it and am not sure what to check for.

so I think Ill need less memory with these bulk collects actually and less
problems with re-querying the same data.

only issue is multiple bulk collects. Ill play with it.

Oracle LazyDBA home page