re: 25X more rows That should be 125X more rows. Or not. We can't really tell based on the information provided. From: iggy_fernandez@xxxxxxxxxxx To: jonathan@xxxxxxxxxxxxxxxxxx; oracle-l@xxxxxxxxxxxxx Subject: RE: Exadata Tuning Question+ Date: Fri, 7 Nov 2014 20:59:15 -0800 I take it back. With incomplete information to draw upon, it's a good guess that SERVICE_RELATIONSHIP is a big table. Perhaps SID is the primary key of SERVICE_LOOKUP and, since the query wants a comma-delimited list of RELATED_SID values, perhaps SERVICE_RELATIONSHIP has 25X more rows than SERVICE_LOOKUP. Hence we have 35 million full table scans of a monster table compared to the single scan that is actually needed. Iggy