Thread: Re: query performance, though it was timestamps,maybe just table size?
Henry Drexler wrote: > why would the query time go from 4 minutes to over 50, for an > increase in table rows from 30 million to 65 million? Did the active (frequently referenced) portion of the database go from something which fit in cache to something which didn't? Did any hash table or sort nodes in plans go from fitting in work_mem to spilling to disk? Did any indexes need an extra level in the tree? Did any plans change based on size to something which is less than optimal, suggesting a need to tune the cost factors? -Kevin
On Fri, Nov 30, 2012 at 1:23 PM, Kevin Grittner <kgrittn@mail.com> wrote:
Thank you for the list - I will research those in the manual.
Henry Drexler wrote:Did the active (frequently referenced) portion of the database go
> why would the query time go from 4 minutes to over 50, for an
> increase in table rows from 30 million to 65 million?
from something which fit in cache to something which didn't? Did
any hash table or sort nodes in plans go from fitting in work_mem
to spilling to disk? Did any indexes need an extra level in the
tree? Did any plans change based on size to something which is less
than optimal, suggesting a need to tune the cost factors?
-Kevin
Thank you for the list - I will research those in the manual.