Vivek Khera wrote:
> > "BM" == Bruce Momjian <[EMAIL PROTECTED]> writes:
>
>
> BM> The 32k number is from Tatsuo testing a few years ago.
>
> Can you verify for me that these parameters in postgresql.conf are
> based on the BLCKSZ (ie one buffer is the size of the BLCKSIZ macro):
>
> shared
Vivek Khera wrote:
> > "BM" == Bruce Momjian <[EMAIL PROTECTED]> writes:
>
>
> BM> The 32k number is from Tatsuo testing a few years ago.
>
> Can you verify for me that these parameters in postgresql.conf are
> based on the BLCKSZ (ie one buffer is the size of the BLCKSIZ macro):
>
> shared
On Wed, 3 Sep 2003, Vivek Khera wrote:
> > "SC" == Sean Chittenden <[EMAIL PROTECTED]> writes:
>
> >> I need to step in and do 2 things:
> SC> Thanks for posting that. Let me know if you have any questions while
> SC> doing your testing. I've found that using 16K blocks on FreeBSD
> SC> res
> "BM" == Bruce Momjian <[EMAIL PROTECTED]> writes:
BM> The 32k number is from Tatsuo testing a few years ago.
Can you verify for me that these parameters in postgresql.conf are
based on the BLCKSZ (ie one buffer is the size of the BLCKSIZ macro):
shared_buffers
effective_cache_size
Logica
Vivek Khera wrote:
> > "MGF" == Marc G Fournier <[EMAIL PROTECTED]> writes:
>
> MGF> Just curious, but Bruce(?) mentioned that apparently a 32k block size was
> MGF> found to show a 15% improvement ... care to run one more test? :)
>
>
> Well, it is hard to tell based on my quick and dirty t
> "MGF" == Marc G Fournier <[EMAIL PROTECTED]> writes:
MGF> Just curious, but Bruce(?) mentioned that apparently a 32k block size was
MGF> found to show a 15% improvement ... care to run one more test? :)
Well, it is hard to tell based on my quick and dirty test:
16k page size:
restore time
Marc G. Fournier wrote:
> On Wed, 3 Sep 2003, Bruce Momjian wrote:
>
> > Vivek Khera wrote:
> > > the restore complained often about checkpoints occurring every few
> > > seconds:
> > >
> > > Sep 2 11:57:14 d02 postgres[49721]: [5-1] LOG: checkpoints are occurring too
> > > frequently (15 secon
On Wed, 3 Sep 2003, Bruce Momjian wrote:
> Vivek Khera wrote:
> > the restore complained often about checkpoints occurring every few
> > seconds:
> >
> > Sep 2 11:57:14 d02 postgres[49721]: [5-1] LOG: checkpoints are occurring too
> > frequently (15 seconds apart)
> > Sep 2 11:57:14 d02 post
> I uppercased it because config parameters are uppercased in the
> documentation. Do we mention config parameters in any other error
> messages? Should it be lowercased?
How about changing the hint?
Consider increasing CHECKPOINT_SEGMENTS in your postgresql.conf
signature.asc
Description: Th
Vivek Khera wrote:
> the restore complained often about checkpoints occurring every few
> seconds:
>
> Sep 2 11:57:14 d02 postgres[49721]: [5-1] LOG: checkpoints are occurring too
> frequently (15 seconds apart)
> Sep 2 11:57:14 d02 postgres[49721]: [5-2] HINT: Consider increasing
> CHECKPOI
Just curious, but Bruce(?) mentioned that apparently a 32k block size was
found to show a 15% improvement ... care to run one more test? :)
On Wed, 3 Sep 2003, Vivek Khera wrote:
> Ok... simple tests have completed. Here are some numbers.
>
> FreeBSD 4.8
> PG 7.4b2
> 4GB Ram
> Dual Xeon 2.4GHz
> "SC" == Sean Chittenden <[EMAIL PROTECTED]> writes:
SC> hardware setup, Vivek, would it be possible for you to run a test with
SC> 32K blocks?
Will do. What's another 4 hours... ;-)
I guess I'll halve the buffer size parameters again...
SC> I've started writing a threaded benchmarking pr
> Ok... simple tests have completed. Here are some numbers.
>
> FreeBSD 4.8
> PG 7.4b2
> 4GB Ram
> Dual Xeon 2.4GHz processors
> 14 U320 SCSI disks attached to Dell PERC3/DC RAID controller in RAID 5
> config with 32k stripe size
[snip]
> Then I took the suggestion to update PG's page size to 16
Ok... simple tests have completed. Here are some numbers.
FreeBSD 4.8
PG 7.4b2
4GB Ram
Dual Xeon 2.4GHz processors
14 U320 SCSI disks attached to Dell PERC3/DC RAID controller in RAID 5
config with 32k stripe size
Dump file:
-rw-r--r-- 1 vivek wheel 1646633745 Aug 28 11:01 19-Aug-2003.dump
> "SC" == Sean Chittenden <[EMAIL PROTECTED]> writes:
>> I need to step in and do 2 things:
SC> Thanks for posting that. Let me know if you have any questions while
SC> doing your testing. I've found that using 16K blocks on FreeBSD
SC> results in about an 8% speedup in writes to the databas
15 matches
Mail list logo
