On 11 June 2010 16:29, Leonardo F <m_lists@xxxxxxxx> wrote: > >> Could you please explain the reason to do so many >> partitions? > > > Because otherwise there would be tons of rows in each > partition, and randomly "updating" the index for that many > rows 2000 times per second isn't doable (the indexes > get so big that it would be like writing a multi-GB file > randomly) > >> In case b) you will face a huge overhead related to necessity >> of >> checking all the data in the table every time new index is >> created > > > I would create the table with all the indexes already in; but only > > the index related to the "current timestamp of the inserted row" > would be updated; the others wouldn't be touched. Well the situation is still ambiguous so: Is it possible to provide this table and indexes definitions? And it would be great it you describe the queries you are going to do on this table or just provide the SQL. -- Sergey Konoplev Blog: http://gray-hemp.blogspot.com / Linkedin: http://ru.linkedin.com/in/grayhemp / JID/GTalk: gray.ru@xxxxxxxxx / Skype: gray-hemp / ICQ: 29353802 -- Sent via pgsql-general mailing list (pgsql-general@xxxxxxxxxxxxxx) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-general
