Re: partitioning by primary key
Posted by:
Rick James
Date: August 08, 2009 04:39PM
Is the data inserted effectively in "created" order? If so, "id" values that are close to each other will correspond to "created" values that are close. Therefore, the clustering on the PK that is inherent in InnoDB leads to clustering on created (or nearly so). Hence, PARTITIONing won't improve the clustering.
What's your value of innodb_buffer_pool_size? It should be perhaps 70% of available RAM.
What do your queries look like? Probably you could construct a "summary" table that has daily totals. Then the 'report' queries could use such a table, and be significantly faster.
Subject
Views
Written By
Posted
3120
July 30, 2009 03:36AM
Re: partitioning by primary key
1884
August 08, 2009 04:39PM
Sorry, you can't reply to this topic. It has been closed.
Content reproduced on this site is the property of the respective copyright holders.
It is not reviewed in advance by Oracle and does not necessarily represent the opinion
of Oracle or any other party.