* Steve Wolfe <steve@iboats.com> [010123 13:11] wrote:
> > Shouldn't it be possible to build vacuum as an ongoing internal PG
> process,
> > instead of a seperate operation? How does Oracle byepass this? Must be
> some
> > way that can be implemented.
>
> Well, here's what it comes down to: Do you want updates to happen
> quickly, and vacuum when load is low, or do you want updates to be slow all
> the time? I suppose that there are some sites that can't find two minutes
> per day when updates will block (not selects), but I imagine they're very
> few.
There are some sites where going for more than an hour without a
VACUUM makes response times unnaceptable, and each vacuum can take
20 minutes a run.
Yes, I'm _an_ exception, but I know I'm not the only one out there.
--
-Alfred Perlstein - [bright@wintelcom.net|alfred@freebsd.org]
"I have the heart of a child; I keep it in a jar on my desk."