Requête lente sur une table partionné.

Bonjour,

J'ai la problématique suivante:

Le requête suivante select max(clock) from history_uint where itemid=xxxx AND clock <= xxxxxx est très lente.

Après vérification du plan d'exécution, il s'avère que le planificateur passe dans toutes les tables filles de history_uint.

J'utilise PostgreSQL 9.0.4.

La structure de la table est la suivante :

                        Table "public.history_uint"
 Column |     Type      |          Modifiers          | Storage | Description 
--------+---------------+-----------------------------+---------+-------------
 itemid | bigint        | not null                    | plain   | 
 clock  | integer       | not null default 0          | plain   | 
 value  | numeric(20,0) | not null default 0::numeric | main    | 
 ns     | integer       | not null default 0          | plain   | 
Indexes:
    "history_uint_1_tmp" btree (itemid, clock)
Triggers:
    insert_history_uint_trigger BEFORE INSERT ON history_uint FOR EACH ROW EXECUTE PROCEDURE history_uint_insert_trigger()
Child tables: history_uint_y2013m08d15,
              history_uint_y2014m10d11,
              history_uint_y2014m10d12,
              history_uint_y2014m10d13,
              history_uint_y2014m10d14,
              history_uint_y2014m10d15,
              history_uint_y2014m10d16,
              history_uint_y2014m10d17,
              history_uint_y2014m10d18,
              history_uint_y2014m10d19,
              history_uint_y2014m10d20,
              history_uint_y2014m10d21,
              history_uint_y2014m10d22,
              history_uint_y2014m10d23,
              history_uint_y2014m10d24,
              history_uint_y2014m10d25,
              history_uint_y2014m10d26,
              history_uint_y2014m10d27,
              history_uint_y2014m10d28,
              history_uint_y2014m10d29,
              history_uint_y2014m10d30,
              history_uint_y2014m10d31,
              history_uint_y2014m11d01,
              history_uint_y2014m11d02,
              history_uint_y2014m11d03,
              history_uint_y2014m11d04,
              history_uint_y2014m11d05,
              history_uint_y2014m11d06,
              history_uint_y2014m11d07,
              history_uint_y2014m11d08,
              history_uint_y2014m11d09,
              history_uint_y2014m11d10,
              history_uint_y2014m11d11,
              history_uint_y2014m11d12,
              history_uint_y2014m11d13,
              history_uint_y2014m11d14,
              history_uint_y2014m11d15,
              history_uint_y2014m11d16,
              history_uint_y2014m11d17,
              history_uint_y2014m11d18,
              history_uint_y2014m11d19,
              history_uint_y2014m11d20,
              history_uint_y2014m11d21,
              history_uint_y2014m11d22,
              history_uint_y2014m11d23,
              history_uint_y2014m11d24,
              history_uint_y2014m11d25,
              history_uint_y2014m11d26,
              history_uint_y2014m11d27,
              history_uint_y2014m12d28,
              history_uint_y2014m12d29,
              history_uint_y2014m12d30,
              history_uint_y2014m12d31,
              history_uint_y2015m01d01,
              history_uint_y2015m01d02,
              history_uint_y2015m01d03,
              history_uint_y2015m01d04,
              history_uint_y2015m01d05,
              history_uint_y2015m01d06,
              history_uint_y2015m01d07,
              history_uint_y2015m01d08,
              history_uint_y2015m01d09,
              history_uint_y2015m01d10
Has OIDs: yes

Chacune des table contient environ 12 000 000 de lignes, et il y a une rotation sur 4 mois.

Un explain de la requête est disponible ici : http://ift.tt/1HUE7Hn

Voici la structure des tables history_uint_yxxxxmxxdxx

                   Table "public.history_uint_y2015m01d10"
 Column |     Type      |          Modifiers          | Storage | Description 
--------+---------------+-----------------------------+---------+-------------
 itemid | bigint        | not null                    | plain   | 
 clock  | integer       | not null default 0          | plain   | 
 value  | numeric(20,0) | not null default 0::numeric | main    | 
 ns     | integer       | not null default 0          | plain   | 
Indexes:
    "history_uint_y2015m01d10_1" btree (itemid, clock)
Check constraints:
    "history_uint_y2015m01d10_clock_check" CHECK (clock >= 1420848000 AND clock < 1420934400)
Inherits: history_uint
Has OIDs: yes

Là j'avoue que je sèche totalement et toute aide serai la bienvenue.

Merci.