Re: Dynamic Partial Index
От | gustavo halperin |
---|---|
Тема | Re: Dynamic Partial Index |
Дата | |
Msg-id | 44E58DEF.1060100@gmail.com обсуждение исходный текст |
Ответ на | Re: Dynamic Partial Index (Jeff Davis <pgsql@j-davis.com>) |
Список | pgsql-general |
Jeff Davis wrote: > On Fri, 2006-08-18 at 00:19 +0300, gustavo halperin wrote: > >>> Create an index on the table, and then periodically move records into a >>> separate archive table. >>> >>> Regards, >>> Jeff Dave >>> >> Thanks, but I have a question. If the table is a BIIIIIIG table, use >> your solution is still a good idea ?? What about to create a partial >> INDEX for the really current date (and not using the function >> current_date) and periodically dropped and created it with the current >> day again and again ?? >> >> > > If you continue to drop and recreate an index like that, it will need to > scan the table during the creation of the index. That will take a while > on a large table. > > You might be better off just using a normal index. To search the index > only takes log(n) time. What problem are you currently having with a > normal index? > > The downside of a normal index on a large table is that the index will > grow large and consume space. I think the solution is to move seldom- > accessed records to a separate archive table. That way, you don't ever > have to scan the archive table unless you do a search in the archives. > > Regards, > Jeff Davi OK, I become totally convinced now. Thank you, Gustavo
В списке pgsql-general по дате отправления: