"Kevin Grittner" writes:
> Robert Haas wrote:
>> So should we give up on this patch?
> No, this is not news; just confirmation of the earlier gut feelings
> and less convincing statistics that there is no problem. Tom's
> argument that if there's no slowdown for common cases, preventing
> O(N
I wrote:
> I remember someone else on the thread saying [...]
> it provided better structure for future enhancements.
Found the reference:
http://archives.postgresql.org/pgsql-hackers/2009-08/msg00078.php
This was the email which I thought confirmed that the changes were
worth it, even in
On Fri, Aug 7, 2009 at 3:36 PM, Sam Mason wrote:
> On Fri, Aug 07, 2009 at 03:18:54PM -0400, Robert Haas wrote:
>> On Fri, Aug 7, 2009 at 3:08 PM, Kevin Grittner
>> wrote:
>> > With the 20 samples from that last round of tests, the answer (rounded
>> > to the nearest percent) is 60%, so "probably
Robert Haas wrote:
> So should we give up on this patch?
No, this is not news; just confirmation of the earlier gut feelings
and less convincing statistics that there is no problem. Tom's
argument that if there's no slowdown for common cases, preventing
O(N^2) behavior for extreme cases is c
On Fri, Aug 07, 2009 at 03:18:54PM -0400, Robert Haas wrote:
> On Fri, Aug 7, 2009 at 3:08 PM, Kevin Grittner
> wrote:
> > With the 20 samples from that last round of tests, the answer (rounded
> > to the nearest percent) is 60%, so "probably noise" is a good summary.
>
> So should we give up on
On Fri, Aug 07, 2009 at 02:08:21PM -0500, Kevin Grittner wrote:
> With the 20 samples from that last round of tests, the answer (rounded
> to the nearest percent) is 60%, so "probably noise" is a good summary.
> Combined with the 12 samples from earlier comparable runs with the
> prior version of t
On Fri, Aug 7, 2009 at 3:08 PM, Kevin
Grittner wrote:
> Sam Mason wrote:
>
>> All we're saying is that we're less than 90% confident that there's
>> something "significant" going on. All the fiddling with standard
>> deviations and sample sizes is just easiest way (that I know of)
>> that statist
Sam Mason wrote:
> All we're saying is that we're less than 90% confident that there's
> something "significant" going on. All the fiddling with standard
> deviations and sample sizes is just easiest way (that I know of)
> that statistics currently gives us of determining this more formally
>
On Fri, Aug 07, 2009 at 10:39:19AM -0500, Kevin Grittner wrote:
> Sam Mason wrote:
> > Yes, all that sounds as though you've got it.
>
> Thanks. I read through it carefully a few times, but I was still only
> 80% confident that I had it more-or-less right. ;-)
And which method did you use to
Sam Mason wrote:
> Yes, all that sounds as though you've got it.
Thanks. I read through it carefully a few times, but I was still only
80% confident that I had it more-or-less right. ;-)
That does seem like a good test, with the advantage of being
relatively easy to calculate. Thanks aga
On Fri, Aug 07, 2009 at 10:19:20AM -0500, Kevin Grittner wrote:
> Sam Mason wrote:
>
> > What do people do when testing this? I think I'd look to something
> > like Student's t-test to check for statistical significance. My
> > working would go something like:
> >
> > I assume the variance
Sam Mason wrote:
> What do people do when testing this? I think I'd look to something
> like Student's t-test to check for statistical significance. My
> working would go something like:
>
> I assume the variance is the same because it's being tested on the
> same machine.
>
> samples
On Tue, Aug 04, 2009 at 10:45:52AM -0400, Tom Lane wrote:
> Sam Mason writes:
> > t = 0.54 ((avg1 - avg2) / (stddev * sqrt(2/samples)))
>
> > We then have to choose how certain we want to be that they're actually
> > different, 90% is a reasonably easy level to hit (i.e. one part in ten,
Sam Mason writes:
> t = 0.54 ((avg1 - avg2) / (stddev * sqrt(2/samples)))
> We then have to choose how certain we want to be that they're actually
> different, 90% is a reasonably easy level to hit (i.e. one part in ten,
> with 95% being more commonly quoted). For 20 samples we have 19
On Mon, Aug 03, 2009 at 10:03:47AM -0500, Kevin Grittner wrote:
> That's about 0.52% slower with the patch. Because there was over 10%
> variation in the numbers with the patch, I tried leaving out the four
> highest outliers on both, in case it was the result of some other
> activity on the syste
On Mon, Aug 03, 2009 at 11:21:43AM -0400, Tom Lane wrote:
> "Kevin Grittner" writes:
> > Over the weekend I ran 40 restores of Milwaukee County's production
> > data using Friday's snapshot with and without the patch. I alternated
> > between patched and unpatched. It appears that this latest ve
IIRC daveg was volunteering to do some tests with his own data; maybe
we should wait for those results.
Unfortunately, I've lost access to the client's data which was showing
bad behaviour under the first heuristic.
--
Josh Berkus
PostgreSQL Experts Inc.
www.pgexperts.com
--
Sent via pgsql
"Kevin Grittner" writes:
> Over the weekend I ran 40 restores of Milwaukee County's production
> data using Friday's snapshot with and without the patch. I alternated
> between patched and unpatched. It appears that this latest version is
> slightly slower for our production database on the same
That's about 0.52% slower with the patch. Because there was over 10%
variation in the numbers with the patch, I tried leaving out the four
highest outliers on both, in case it was the result of some other
activity on the system (even though this machine should have been
pretty quiet over the we
I wrote:
> Tom Lane wrote:
>> Attached is a further small improvement that gets rid of the
>> find_ready_items() scans. After re-reading the patch I realized
>> that it wasn't *really* avoiding O(N^2) behavior ... but this
>> version does.
>
> I'll run a fresh set of benchmarks.
Over the
daveg writes:
> Will the patch apply to a vanilla 8.4.0?
Yeah, it should. The line numbers in the version I just posted might
be off a little bit for 8.4.0, but patch should cope.
Be sure to "make clean" and recompile all of src/bin/pg_dump, else you
might have some issues.
On Thu, Jul 30, 2009 at 12:29:34PM -0500, Kevin Grittner wrote:
> Tom Lane wrote:
>
> > I think we've pretty much established that it doesn't make things
> > *worse*, so I'm sort of inclined to go ahead and apply it. The
> > theoretical advantage of eliminating O(N^2) search behavior seems
> >
Tom Lane wrote:
> "Kevin Grittner" writes:
>> Rebased to correct for pg_indent changes.
>
> Thanks for doing that.
No problem. I think I still owe you a few. :-)
> Attached is a further small improvement that gets rid of the
> find_ready_items() scans. After re-reading the patch I realiz
"Kevin Grittner" writes:
> Rebased to correct for pg_indent changes.
Thanks for doing that. Attached is a further small improvement that
gets rid of the find_ready_items() scans. After re-reading the patch
I realized that it wasn't *really* avoiding O(N^2) behavior ... but
this version does.
Tom Lane wrote:
> I think we've pretty much established that it doesn't make things
> *worse*, so I'm sort of inclined to go ahead and apply it. The
> theoretical advantage of eliminating O(N^2) search behavior seems
> like reason enough, even if it takes a ridiculous number of tables
> for th
On Thu, Jul 30, 2009 at 1:24 PM, Tom Lane wrote:
> "Kevin Grittner" writes:
>> The timings vary by up to 2.5% between runs, so that's the noise
>> level. Five runs of each (alternating between the two) last night
>> give an average performance of 1.89% faster for the patched version.
>> Combining
"Kevin Grittner" writes:
> The timings vary by up to 2.5% between runs, so that's the noise
> level. Five runs of each (alternating between the two) last night
> give an average performance of 1.89% faster for the patched version.
> Combining that with yesterday's results starts to give me prett
"Kevin Grittner" wrote:
> with the default settings, the patched version ran an additional 1%
> faster than the unpatched; although I don't have enough samples to
> have a high degree of confidence it wasn't noise. I'll run another
> slew of tests tonight with the existing dump file to confirm
"Kevin Grittner" writes:
> Tom Lane wrote:
>> Also, the followup to that message points out that the 8.4.0 code
>> has a potential O(N^2) dependency on the total number of TOC items
>> in the dump. So it might be interesting to check the behavior with
>> very large numbers of tables/indexes.
Tom Lane wrote:
> Also, the followup to that message points out that the 8.4.0 code
> has a potential O(N^2) dependency on the total number of TOC items
> in the dump. So it might be interesting to check the behavior with
> very large numbers of tables/indexes.
I've got 431 user tables with
Robert Haas wrote:
> This is what I've been able to find on a quick look:
>
> http://archives.postgresql.org/pgsql-hackers/2009-05/msg00678.php
>
> Sounds like Kevin may want to try renaming some of his indices to
> produce intermingling...
Thanks, I'll give that a try. Renaming them is on
Robert Haas writes:
> On Tue, Jul 28, 2009 at 9:52 PM, Tom Lane wrote:
>> I don't have time to look right now, but ISTM the original discussion
>> that led to making that patch had ideas about scenarios where it would
>> be faster.
> This is what I've been able to find on a quick look:
> http://a
On Tue, Jul 28, 2009 at 9:52 PM, Tom Lane wrote:
> Robert Haas writes:
>> The other possibility here is that this just doesn't work. :-)
>
> That's why we wanted to test it ;-).
>
> I don't have time to look right now, but ISTM the original discussion
> that led to making that patch had ideas abo
Robert Haas writes:
> The other possibility here is that this just doesn't work. :-)
That's why we wanted to test it ;-).
I don't have time to look right now, but ISTM the original discussion
that led to making that patch had ideas about scenarios where it would
be faster. It'd be worth diggin
On Tue, Jul 28, 2009 at 10:28 AM, Kevin
Grittner wrote:
> I wrote:
>
>> So far, all tests have shown no difference in performance based on
>> the patch;
>
> My testing to that point had been on a "big" machine with 16 CPUs and
> 128 GB RAM and dozens of spindles. Last night I tried with a dual
> c
I wrote:
> So far, all tests have shown no difference in performance based on
> the patch;
My testing to that point had been on a "big" machine with 16 CPUs and
128 GB RAM and dozens of spindles. Last night I tried with a dual
core machine with 4 GB RAM and 5 spindles in RAID 5. Still no
dif
Andrew Dunstan wrote:
> Does your test case have lots of foreign keys?
488 of them.
There is some variation on individual tests, but the results look to
be "in the noise." When I add them all up, the patch comes out
0.0036% slower -- but that is so far into the noise as to be
considered "no
Kevin Grittner wrote:
Andrew Dunstan wrote:
To performance test this properly you might need to devise a test
that will use a sufficiently different order of queueing items to
show the difference.
It would appear that I need help with devising a proper test. So far,
all tests ha
Andrew Dunstan wrote:
> To performance test this properly you might need to devise a test
> that will use a sufficiently different order of queueing items to
> show the difference.
It would appear that I need help with devising a proper test. So far,
all tests have shown no difference in perf
Stefan Kaltenbrunner wrote:
>> My plan here would be to have
>> the dump on one machine, and run pg_restore there, and push it to a
>> database on another machine through the LAN on a 1Gb connection.
>> (This seems most likely to be what we'd be doing in real life.)
> you need to be careful he
Robert Haas wrote:
> it might be worth testing with default settings too.
OK. I'll do that too, if time allows.
-Kevin
--
Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers
Andrew Dunstan wrote:
> To performance test this properly you might need to devise a test
> that will use a sufficiently different order of queueing items to
> show the difference.
>
> One thing I am particularly interested in is to see if queuing FK
> items for a table as soon as they become a
Kevin,
It would be hard to schedule the requisite time on our biggest web
machines, but I assume an 8 core 64GB machine would give meaningful
results. Any sense what numbers of parallel jobs I should use for
tests? I would be tempted to try 1 (with the -1 switch), 8, 12, and
16 -- maybe keep g
Robert Haas wrote:
On Sat, Jul 18, 2009 at 4:41 PM, Kevin
Grittner wrote:
"Kevin Grittner" wrote:
Performance tests to follow in a day or two.
I'm looking to beg another week or so on this to run more tests. What
I can have by the end of today is pretty limited, mostly beca
Kevin Grittner wrote:
"Kevin Grittner" wrote:
Performance tests to follow in a day or two.
I'm looking to beg another week or so on this to run more tests. What
I can have by the end of today is pretty limited, mostly because I
decided it made the most sense to test this with big complex
On Sat, Jul 18, 2009 at 4:41 PM, Kevin
Grittner wrote:
> "Kevin Grittner" wrote:
>
>> Performance tests to follow in a day or two.
>
> I'm looking to beg another week or so on this to run more tests. What
> I can have by the end of today is pretty limited, mostly because I
> decided it made the m
"Kevin Grittner" wrote:
> Performance tests to follow in a day or two.
I'm looking to beg another week or so on this to run more tests. What
I can have by the end of today is pretty limited, mostly because I
decided it made the most sense to test this with big complex
databases, and it just
Rebased to correct for pg_indent changes.
Applies cleanly.
Compiles cleanly.
Passes regression tests.
Comments and format look good.
No documentation changes needed.
No regression test changes needed.
Performance tests to follow in a day or two.
-Kevin
Index: src/bin/pg_dump/pg_backup_archive
48 matches
Mail list logo