Recommendations for partitioning?
От | Dave Johansen |
---|---|
Тема | Recommendations for partitioning? |
Дата | |
Msg-id | CAAcYxUemBq7mqv_-1BVPe36udVEJjDoF2_-9G4Azv7nXZtmZ3A@mail.gmail.com обсуждение исходный текст |
Ответы |
Re: Recommendations for partitioning?
Re: Recommendations for partitioning? |
Список | pgsql-performance |
I'm managing a database that is adding about 10-20M records per day to a table and time is a core part of most queries, so I've been looking into seeing if I need to start using partitioning based on the time column and I've found these general guidelines:
Don't use more than about 50 paritions ( http://www.postgresql.org/message-id/17147.1271696670@sss.pgh.pa.us ) Use triggers to make the interface easier ( https://wiki.postgresql.org/wiki/Table_partitioning#Trigger-based and http://stackoverflow.com/questions/16049396/postgres-partition-by-week )
The only data I found fell inline with what you'd expect (i.e. speeds up selects but slows down inserts/updates http://www.if-not-true-then-false.com/2009/performance-testing-between-partitioned-and-non-partitioned-postgresql-tables-part-3/ )
So I was thinking that partitioning based on month to keep the number of partitions low, so that would mean about 0.5G records in each table. Does that seem like a reasonable number of records in each partition? Is there anything else that I should consider or be aware of?
Thanks,
Thanks,
Dave
В списке pgsql-performance по дате отправления: