Our 32gb solaris 9 server running two instances of 10gr2 One small db to support OID (sga max size 288M, pga aggregate 32M) and one of our production dbs(6gb pga aggregate target, 12gb sga_max_size). According to top, before the workday starts, we are sitting at 10-12gb available physical memory. Once our users get working, that number drops to 500-600mb available. Top reports 29G swap in use, 11G swap free. OEM says that we're performing heavy swapping and I'm not sure how to resolve this. We typically have 1500-1700 connections. My gut tells me that pga might be too small and that sga is too large, but i'm not really able to tell. My v$pgastat and v$sgastat can be viewed at http://spreadsheets.google.com/ccc?key=0ArUTHdqN0J--dFZBNE0wU0JpUW1Na1hVYlBfeFVYOEE&hl=en