Thread: scaling

scaling

From
"Scott V. McGuire"
Date:
Hi,

I'm surprised about how long certain things are taking.  Here's what
I'm doing:

1. create a temp_table which is 5-7 columns by 1e4 - 1e5 rows.
This part is pretty quick.

2. when its done, copy it into the big table which is 7 columns by
about 5e7 rows.

What I'm surprised about is that the 10,000 row copies take < 1
minute while the 50000 row copies take > 10 minutes.

I don't really know what information is relevant here, so tell me and
I'll provide any other details.  (I'd describe the whole application
but didn't want to waste peoples time.)

Thanks,
--
Scott V. McGuire <svmcguir@syr.edu>
GnuPG key available at http://physics.syr.edu/~svmcguir
GnuPG key fingerprint: 21EA 4999 3620 3E1D 71EC  98A9 5B9B EF52 1258 6D53
GnuPG is at http://www.gnupg.org/


Re: [GENERAL] scaling

From
Ed Loehr
Date:
"Scott V. McGuire" wrote:
>
> I'm surprised about how long certain things are taking...
>
> What I'm surprised about is that the 10,000 row copies take < 1
> minute while the 50000 row copies take > 10 minutes.

See http://www.deja.com/getdoc.xp?AN=563958392

I suspect vacuum also would have a dramatic impact on performance in
your test case.

Cheers,
Ed Loehr

Re: [GENERAL] scaling

From
"Scott V. McGuire"
Date:
On Wed, Jan 26, 2000 at 04:51:26PM -0600, Ed Loehr wrote:
> See http://www.deja.com/getdoc.xp?AN=563958392
>
> I suspect vacuum also would have a dramatic impact on performance in
> your test case.
>
> Cheers,
> Ed Loehr
>

Ok, well there's at least two things I should have said.  I am running
with -F and I did a vacuum analyze last night.  Does vacuum analyze do
the analyze stuff in addition to or instead of the plain vacuum stuff?

--
Scott V. McGuire <svmcguir@syr.edu>
GnuPG key available at http://physics.syr.edu/~svmcguir
GnuPG key fingerprint: 21EA 4999 3620 3E1D 71EC  98A9 5B9B EF52 1258 6D53
GnuPG is at http://www.gnupg.org/


Re: [GENERAL] scaling

From
Ed Loehr
Date:
"Scott V. McGuire" wrote:
>
> > See http://www.deja.com/getdoc.xp?AN=563958392
> >
> > I suspect vacuum also would have a dramatic impact on performance in
> > your test case.
>
> Ok, well there's at least two things I should have said.  I am running
> with -F and I did a vacuum analyze last night.  Does vacuum analyze do
> the analyze stuff in addition to or instead of the plain vacuum stuff?

In addition, I believe.

It also occurred to me that it probably doesn't really matter how much
*time* has passed since your last vacuum, but rather how much
*activity* (ie., inserts/deletes) has occurred since you last
vacuumed...

Cheers,
Ed Loehr