On Sun, 2009-06-07 at 12:13 -0400, Tom Lane wrote:
> In any case, what we seem to have here is evidence that there are some
> cases where the new default value of default_statistics_target is too
> high and you can get a benefit by lowering it. I'm not sure we should
> panic about that. Default
On Sun, Jun 7, 2009 at 5:13 PM, Tom Lane wrote:
>
> In any case, what we seem to have here is evidence that there are some
> cases where the new default value of default_statistics_target is too
> high and you can get a benefit by lowering it. I'm not sure we should
> panic about that. Default va
Simon Riggs writes:
> On Sat, 2009-06-06 at 12:06 -0700, Josh Berkus wrote:
>> Well, Jignesh and I identified two things which we think are "special"
>> about DBT2: (1) it uses C stored procedures, and (2) we don't think it
>> uses prepared plans.
> If there is a performance regression it is al
On Sat, 2009-06-06 at 12:06 -0700, Josh Berkus wrote:
> > On the DL380 GB system, where I'm using a lot more drives the Jignesh,
> > I see a performance change of under 5%. 15651.14 notpm vs 16333.32
> > notpm. And this is after a bit of tuning, not sure how much the out
> > of the box experien
Mark,
On the DL380 GB system, where I'm using a lot more drives the Jignesh,
I see a performance change of under 5%. 15651.14 notpm vs 16333.32
notpm. And this is after a bit of tuning, not sure how much the out
of the box experience changes on this system.
Well, Jignesh and I identified two
On Fri, May 22, 2009 at 10:38 AM, Tom Lane wrote:
> Greg Smith writes:
>> Yesterday Jignesh Shah presented his extensive benchmark results comparing
>> 8.4-beta1 with 8.3.7 at PGCon:
>> http://blogs.sun.com/jkshah/entry/pgcon_2009_performance_comparison_of
>
>> While most cases were dead even or a
On 22 May 2009, at 16:17, Greg Smith wrote:
On Fri, 22 May 2009, Greg Sabino Mullane wrote:
The bump from 10 to 100 was supported by microbenchmarks that
suggested it
would be tolerable.
No, the 10 to 100 was supported by years of people working in the
field who
routinely did that ad
On Fri, 22 May 2009, Greg Sabino Mullane wrote:
The bump from 10 to 100 was supported by microbenchmarks that suggested it
would be tolerable.
No, the 10 to 100 was supported by years of people working in the field who
routinely did that adjustment (and >100) and saw great gains.
No one is s
On 5/22/09 2:36 PM, Tom Lane wrote:
and...@dunslane.net writes:
Wouldn't he just need to rerun the tests with default_stats_target set to
the old value? I presume he has actually done this already in order to
come to the conclusion he did about the cause of the regression.
Yeah, he did, so we
On Fri, May 22, 2009 2:41 pm, Tom Lane wrote:
> "Greg Sabino Mullane" writes:
>> No, the 10 to 100 was supported by years of people working in the
>> field who routinely did that adjustment (and >100) and saw great
>> gains. Also, as the one who originally started the push to 100, my
>> original g
"Greg Sabino Mullane" writes:
> No, the 10 to 100 was supported by years of people working in the
> field who routinely did that adjustment (and >100) and saw great
> gains. Also, as the one who originally started the push to 100, my
> original goal was to get it over the "magic 99" bump, at which
"Joshua D. Drake" writes:
> We probably need to test this to get some more data points.
Agreed --- DBT2 is just one data point. We shouldn't assume that it's
definitive.
regards, tom lane
--
Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org)
To make cha
and...@dunslane.net writes:
> Wouldn't he just need to rerun the tests with default_stats_target set to
> the old value? I presume he has actually done this already in order to
> come to the conclusion he did about the cause of the regression.
Yeah, he did, so we know it's slower that way. But ex
> Greg Smith writes:
>> Yesterday Jignesh Shah presented his extensive benchmark results
>> comparing
>> 8.4-beta1 with 8.3.7 at PGCon:
>> http://blogs.sun.com/jkshah/entry/pgcon_2009_performance_comparison_of
>
>> While most cases were dead even or a modest improvement, his dbt-2
>> results
>> su
Greg Sabino Mullane wrote:
-BEGIN PGP SIGNED MESSAGE-
Hash: RIPEMD160
While most cases were dead even or a modest improvement, his dbt-2 results
suggest a 15-20% regression in 8.4. Changing the default_statistics_taget
to 100 was responsible for about 80% of that regression.
On Fri, 2009-05-22 at 13:35 -0400, Stephen Frost wrote:
> Greg,
> dbt-2 is for OLTP, not for DW. Greg Smith's comment was actually that
> we shouldn't penalize the OLTP crowd (by raising the value) for the
> benefit of the DW crowd (who need it higher than 100 anyway).
>
I appear to have comple
Greg Smith writes:
> Yesterday Jignesh Shah presented his extensive benchmark results comparing
> 8.4-beta1 with 8.3.7 at PGCon:
> http://blogs.sun.com/jkshah/entry/pgcon_2009_performance_comparison_of
> While most cases were dead even or a modest improvement, his dbt-2 results
> suggest a 15-
Greg,
* Greg Sabino Mullane (g...@turnstep.com) wrote:
> > While most cases were dead even or a modest improvement, his dbt-2 results
> > suggest a 15-20% regression in 8.4. Changing the default_statistics_taget
> > to 100 was responsible for about 80% of that regression.
> ...
> > The situation
On Fri, 2009-05-22 at 16:43 +, Greg Sabino Mullane wrote:
> -BEGIN PGP SIGNED MESSAGE-
> Hash: RIPEMD160
>
>
> > While most cases were dead even or a modest improvement, his dbt-2 results
> > suggest a 15-20% regression in 8.4. Changing the default_statistics_taget
> > to 100 was res
-BEGIN PGP SIGNED MESSAGE-
Hash: RIPEMD160
> While most cases were dead even or a modest improvement, his dbt-2 results
> suggest a 15-20% regression in 8.4. Changing the default_statistics_taget
> to 100 was responsible for about 80% of that regression.
...
> The situation where the st
Yesterday Jignesh Shah presented his extensive benchmark results comparing
8.4-beta1 with 8.3.7 at PGCon:
http://blogs.sun.com/jkshah/entry/pgcon_2009_performance_comparison_of
While most cases were dead even or a modest improvement, his dbt-2 results
suggest a 15-20% regression in 8.4. Chang
21 matches
Mail list logo
