Guys,
Have 20Gb and 16CPUS available on host. Need to do large full-table scans/joins to create materialized view. Since I have to do the full-table scans of large tables - decided to use parallel query option. Eliminated significant I/O contention by using DIRECT IO. Using very very large sort_area_size, however, still writing out significant segments to temporary tablespace which doesn't make sense to me. Any ideas?