We noticed that after upgrading to 7.1beta the execution time for
some often used queries went up by a factor of 2 or more. Considering
the early beta state I was not alarmed. But since I noticed that
yesterday's snapshot still has the problem, I'd really like to tell you
Here is one of the queries, it takes about half a second on our computer
(PII 233 with 256MB RAM) to execute and returns typically 1-4 rows via
two index scans with high selectivity. So it looks to me that planning
time outwages execution time by far. 7.0 took about 0.15 seconds (which
is still much).
Here is the query:
explain verbose select gaenge , s . artikelid , text from
schaertabelle s , extartbez e where maschine = int2(109) and
schaerdatum = '2001-01-13' and s . artikelid = e . artikelid and
extartbezid = 1 and bezkomptype = 0 order by text limit 10;
And the plan for 7.0 and 7.1 (attached).
The data and schema is accessible via
If you omit 'int2(' the index scan collapses into a sequential scan.
(Well known problem with int2 indices)
Oh, I'll attach the schema, too. So if you just want to take a look at
the table definition you don't have to download the data.
Description: text/plain (24.5 KB)
Description: text/plain (18.9 KB)
Description: text/plain (1.2 KB)
pgsql-hackers by date
|Next:||From: Vic||Date: 2001-03-05 08:33:35|
|Subject: Oops! Its bug in parser???? |
|Previous:||From: Patrick Dunford||Date: 2001-03-05 07:11:28|
|Subject: Getting unique ID through SQL|