Re: Re: Planner estimates cost of 'like' a lot lower than '='??

Precisely the point! There "shouldn't" be an issue with SMP boxes other
than OS differences that may result in one OS more efficiently utilizing
additional processors, yet there were striking differences in the
execution of my queries when I dropped the index after moving my system
from a single processor to a dual processor box using the identical
software configuration.

At the time, I *thought* it may have been a PostgreSQL issue, but it
just as well may have been a Linux issue. Your similar situation on BSD
leads me to believe that it may in fact be an issue with PostgreSQL
after all.

Disturbing isn't it? ;)

Mats Lofkvist wrote:
>
> haystack(at)email(dot)rentzone(dot)org (Buddy Lee Haystack) writes:
>
> > I ran into a similar problem on RedHat Linux v6.1 on Intel, kernel
> > 2.2.12-20, PostgreSQL 6.5.3 when moving a system from a single processor
> > development box over to a dual processor production server. Dropping one
> > of the indexes on a lookup table with roughly 68,000 records on the
> > production box resulted in roughly a 3 fold increase in query execution
> > speed. At the time, I thought it was an SMP issue, and have since been
> > extremely conservative in adding indexes on SMP boxes.
>
> I _am_ running it on an SMP box (FreeBSD 5.0-current from january
> this year), but isn't it a bit far-fetched to assume that this is
> an SMP issue? Is postgres even aware of running on an SMP box?
> (if it isn't, why should the planner estimates differ depending
> on if it is running on an SMP box or not?)
>
> Forgive me for sounding negative, but I fail to see the connection.
> Am I missing something?
>
> _
> Mats Lofkvist
> mal(at)algonet(dot)se
>
> ---------------------------(end of broadcast)---------------------------
> TIP 2: you can get off all lists at once with the unregister command
> (send "unregister YourEmailAddressHere" to majordomo(at)postgresql(dot)org)