Обсуждение: different query plan for same select

Поиск
Список
Период
Сортировка

different query plan for same select

От
Jörg Schulz
Дата:
I have these two tables with the same data on two different
machines (SuSE 8.2 and Gentoo both with 7.3.2):


schulz=> \d rechnung
       Table "jschulz.rechnung"
     Column     |   Type   | Modifiers
----------------+----------+-----------
 system         | smallint | not null
 jahr           | smallint | not null
 monat          | smallint | not null
 rechnungsnr    | integer  | not null
 rechnungsdatum | date     | not null
 kundennr       | integer  | not null
 seiten         | smallint | not null
 formularnr     | smallint |
 text           | text     | not null
Indexes: rechnung_pkey primary key btree (system, jahr, rechnungsnr),
         rechnung_kundennr btree (kundennr),
         rechnung_rechnungsdatum btree (rechnungsdatum),
         rechnung_rechnungsnr btree (rechnungsnr)

schulz=> \d rechnung_zusatz
  Table "jschulz.rechnung_zusatz"
   Column    |   Type   | Modifiers
-------------+----------+-----------
 system      | smallint | not null
 jahr        | smallint | not null
 rechnungsnr | integer  | not null
 objektnr    | integer  |
Indexes: rechnung_zusatz_uniq_objektnr unique btree (system, jahr,
rechnungsnr, objektnr),
         rechnung_zusatz_objektnr btree (objektnr)
Foreign Key constraints: $1 FOREIGN KEY (system, jahr, rechnungsnr) REFERENCES
rechnung(system, jahr, rechnungsnr) ON UPDATE NO ACTION ON DELETE NO ACTION

schulz=>


On the SuSE machine an explain gives the following:


schulz=> explain select system, jahr, rechnungsnr from (rechnung natural left
join rechnung_zusatz) where objektnr=1;
                                      QUERY PLAN
--------------------------------------------------------------------------------------
 Hash Join  (cost=0.00..25.04 rows=1000 width=20)
   Hash Cond: ("outer".rechnungsnr = "inner".rechnungsnr)
   Join Filter: (("outer".system = "inner".system) AND ("outer".jahr =
"inner".jahr))
   Filter: ("inner".objektnr = 1)
   ->  Seq Scan on rechnung  (cost=0.00..20.00 rows=1000 width=8)
   ->  Hash  (cost=0.00..0.00 rows=1 width=12)
         ->  Seq Scan on rechnung_zusatz  (cost=0.00..0.00 rows=1 width=12)
(7 rows)

schulz=>


On the Gentoo machine the same explain gives:


schulz=> explain select system, jahr, rechnungsnr from (rechnung natural left
join rechnung_zusatz) where objektnr=1;
                                                             QUERY PLAN
             
---------------------------------------------------------------------------------
 Merge Join  (cost=0.00..109.00 rows=1000 width=20)
   Merge Cond: (("outer".system = "inner".system) AND ("outer".jahr =
"inner".jahr) AND ("outer".rechnungsnr = "inner".rechnungsnr))
   Filter: ("inner".objektnr = 1)
   ->  Index Scan using rechnung_pkey on rechnung  (cost=0.00..52.00 rows=1000
width=8)
   ->  Index Scan using rechnung_zusatz_uni_objektnr on rechnung_zusatz
(cost=0.00..52.00 rows=1000 width=12)
(5 Zeilen)

schulz=>


The select on the SuSE machine finishes in about 3 seconds and on the
Gentoo machine it doesn't seem to come to an end at all. Each table has
about 80.000 rows.

I'm not very familar with the output of the explain command but can you
tell me why I get two different query plans?


Jörg

Re: different query plan for same select

От
Tom Lane
Дата:
=?iso-8859-1?q?J=F6rg=20Schulz?= <jschulz@sgbs.de> writes:
> I'm not very familar with the output of the explain command but can you
> tell me why I get two different query plans?

Judging from the suspiciously round numbers in the cost estimates,
you've never done a VACUUM ANALYZE on any of these tables.  Try that and
then see what plans you get...

            regards, tom lane

Re: different query plan for same select

От
Jörg Schulz
Дата:
> > I'm not very familar with the output of the explain command but can you
> > tell me why I get two different query plans?
>
> Judging from the suspiciously round numbers in the cost estimates,
> you've never done a VACUUM ANALYZE on any of these tables.  Try that and
> then see what plans you get...

Oops.. You were right!

Thank you anyway.

Jörg