Re: Choice of bitmap scan over index scan
От | Kevin Grittner |
---|---|
Тема | Re: Choice of bitmap scan over index scan |
Дата | |
Msg-id | 4B49A3AF020000250002E138@gw.wicourts.gov обсуждение исходный текст |
Ответ на | Choice of bitmap scan over index scan (Mathieu De Zutter <mathieu@dezutter.org>) |
Ответы |
Re: Choice of bitmap scan over index scan
|
Список | pgsql-performance |
Mathieu De Zutter wrote: > Intel(R) Core(TM)2 Duo CPU E7200 @ 2.53GHz > 2GB RAM > 2x500GB RAID-1 > Running Debian/Etch AMD64 > PG version: PostgreSQL 8.3.8 on x86_64 > Server also runs DNS/Mail/Web/VCS/... for budget reasons. > Database size is 1-2 GB. Also running copies of it for testing/dev. I would try something like this and see how it goes: shared_buffers = 200MB work_mem = 32MB effective_cache_size = 1.2GB seq_page_cost = 0.1 random_page_cost = 0.1 Some of these settings require a PostgreSQL restart. I may have gone too aggressively low on the page costs, but it seems likely with a "1-2 GB" database and 2 GB RAM, the active portions of your database will be fully cached in spite of the other applications. Besides looking at the impact on this one query, you should keep an eye on all queries after such changes, and post for any which become unacceptably slow. Properly tuning something like this can take a few iterations. -Kevin
В списке pgsql-performance по дате отправления: