Tom Lane <tgl@sss.pgh.pa.us> wrote:=0A=0A> I'm wondering about cache effect=
s, ie memory already contains=0A> desired pages in the morning (perhaps as =
a side-effect of queries=0A> run overnight) and not so much by the evening.=
=A0 If so it's not=0A> clear that additional VACUUM activity would make thi=
ngs better.=0A>=0A> But in any case it's hard to diagnose this without EXPL=
AIN=0A> ANALYZE output.=0A=0AHe did show EXPLAIN OUTPUT results, which show=
ed an index scan in=0Athe morning and a seqscan in the evening.=A0 He also =
mentioned that=0Ahe doubted cache effects amounted to much because the data=
base was=0A40GB and the machine has 64GB.=0A=0Ahttp://www.postgresql.org/me=
ssage-id/CAJnnue1x6YYqKbvDQkMTonK62GZSPH-edPR4U7+LOt66M3s_Sw@mail.gmail.com=
=0A=0A-- =0AKevin Grittner=0AEnterpriseDB: http://www.enterprisedb.com=0ATh=
e Enterprise PostgreSQL Company