Received: from localhost.localdomain (postgresql.org [64.49.215.8]) by localhost (Postfix) with ESMTP id 30A3F476728 for ; Thu, 2 May 2002 10:15:23 -0400 (EDT) Received: from sss.pgh.pa.us (unknown [192.204.191.242]) by postgresql.org (Postfix) with ESMTP id AD78A476313 for ; Thu, 2 May 2002 09:58:29 -0400 (EDT) Received: from sss2.sss.pgh.pa.us (tgl@localhost [127.0.0.1]) by sss.pgh.pa.us (8.11.4/8.11.4) with ESMTP id g42DwCv07119; Thu, 2 May 2002 09:58:12 -0400 (EDT) To: Wolfgang.Fuertbauer@ebewe.com Cc: pgsql-sql@postgresql.org, Stephan Szabo Subject: Re: join-performance problem In-reply-to: References: Comments: In-reply-to Wolfgang.Fuertbauer@ebewe.com message dated "Thu, 02 May 2002 10:00:47 +0200" Date: Thu, 02 May 2002 09:58:12 -0400 Message-ID: <7116.1020347892@sss.pgh.pa.us> From: Tom Lane X-Archive-Number: 200205/12 X-Sequence-Number: 7648 Wolfgang.Fuertbauer@ebewe.com writes: > Hash Join (cost=482.82..704.62 rows=80 width=93) > -> Hash Join (cost=481.78..702.17 rows=80 width=79) > -> Hash Join (cost=480.44..699.43 rows=80 width=59) > -> Seq Scan on artikel c (cost=0.00..155.77 rows=4977 > width=18) > -> Hash (cost=480.24..480.24 rows=80 width=41) > -> Nested Loop (cost=0.00..480.24 rows=80 width=41) > -> Index Scan using fakt_kunde_key on fakt b > (cost=0.00..55.11 rows=13 width=16) > -> Index Scan using faktzeilen_pkey on > faktzeilen a (cost=0.00..31.82 rows=10 width=25) > -> Hash (cost=1.27..1.27 rows=27 width=20) > -> Seq Scan on argruppen e (cost=0.00..1.27 rows=27 > width=20) > -> Hash (cost=1.03..1.03 rows=3 width=14) > -> Seq Scan on kollektion d (cost=0.00..1.03 rows=3 width=14) > can you pleas explain me why c is now scanned sequential? Looks like a reasonable choice to me. 80 indexscan probes would probably cost more than the one seqscan, given that c is so small. The alternative of putting c on the inside of the hashjoin won't be chosen as long as c is the larger relation... regards, tom lane