Re: Performance regression between 8.3 and 8.4 on heavy text indexing

From: Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us>
To: gael(at)pilotsystems(dot)net ( Gaël Le Mignot)
Cc: Guillaume Smet <guillaume(dot)smet(at)gmail(dot)com>, pgsql-performance(at)postgresql(dot)org
Subject: Re: Performance regression between 8.3 and 8.4 on heavy text indexing
Date: 2009-08-26 16:29:45
Message-ID: 17025.1251304185@sss.pgh.pa.us
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-performance

gael(at)pilotsystems(dot)net (=?iso-8859-1?Q?Ga=EBl?= Le Mignot) writes:
> So it seems it was quite wrong about estimated matching rows (192 predicted, 10222 reals).

Yup. What's even more interesting is that it seems the real win would
have been to use just the 'claude & duviau' condition (which apparently
matched only 14 rows). 8.3 had no hope whatever of understanding that,
it just got lucky. 8.4 should have figured it out, I'm thinking.
Does it help if you increase the statistics target for fulltext_tsv?
(Don't forget to re-ANALYZE after doing so.)

regards, tom lane

In response to

Responses

Browse pgsql-performance by date

  From Date Subject
Next Message Hrishikesh (हृषीकेश मेहेंदळे) 2009-08-26 17:31:13 Performance issues with large amounts of time-series data
Previous Message Gaël Le Mignot 2009-08-26 16:03:34 Re: Performance regression between 8.3 and 8.4 on heavy text indexing