Re: Index location patch for review (more pgbench results)

Started by Jim Buttafuocoover 24 years ago2 messageshackers
Jump to latest
#1Jim Buttafuoco
jim@buttafuoco.net

Moving the test to a system with SCSI disks gave different results.
There is NO difference between having the indexes on the same disk or
different disk with the data while running pgbench. So I leave it up to
you guys as to include the patch or not. I do believe that even if
performance doesn't increase, this patch as alot of other benefits for
admins.

Let me know
Jim

Here is my pgbench results. As you can see the I am getting 2X tps

with

the 2 directories. I believe this is a BIG win for Postgresql if we

can

figure out the WAL recovery issues.

Can someone other than me apply the patch and verify the pgbench
results.

My hardward setup is a dual processor P3/733 running Redhat 7.1 with

512

megs of memory. The postgresql.conf file is the installed version with
NO changes.

Jim

template1=# create database one_dir with location='PGDATA1';
template1=# create database two_dir with location='PGDATA1'
index_location='PGIDX1';
for X in 1 2 3 4 5 6 7 8 9 10
do
pgbench -i -s 10 one_dir >>one_dir.log
pgbench -i -s 10 two_dir >>two_dir.log
done

bash-2.04$ grep 'excluding' one_dir.log
tps = 44.319306(excluding connections establishing)
tps = 34.641020(excluding connections establishing)
tps = 50.516889(excluding connections establishing)
tps = 52.747039(excluding connections establishing)
tps = 16.203821(excluding connections establishing)
tps = 36.902861(excluding connections establishing)
tps = 52.511769(excluding connections establishing)
tps = 53.479882(excluding connections establishing)
tps = 54.599429(excluding connections establishing)
tps = 36.780419(excluding connections establishing)
tps = 48.048279(excluding connections establishing)

bash-2.04$ grep 'excluding' two_dir.log
tps = 58.739049(excluding connections establishing)
tps = 100.259270(excluding connections establishing)
tps = 103.156166(excluding connections establishing)
tps = 110.829358(excluding connections establishing)
tps = 111.929690(excluding connections establishing)
tps = 106.840118(excluding connections establishing)
tps = 101.563159(excluding connections establishing)
tps = 102.877060(excluding connections establishing)
tps = 103.784717(excluding connections establishing)
tps = 53.056309(excluding connections establishing)
tps = 73.842428(excluding connections establishing)

Also I have been running this patch (both 7.1.3 and 7.2devel)

against

some of my companies applications. I have loaded a small database

10G

We are not familiar with your applications. It would be better to

see

Show quoted text

results of test suit available to the community. pgbench is first to
come in mind. Such tests would be more valuable.

Vadim

---------------------------(end of

broadcast)---------------------------

TIP 4: Don't 'kill -9' the postmaster

#2Bruce Momjian
bruce@momjian.us
In reply to: Jim Buttafuoco (#1)

Moving the test to a system with SCSI disks gave different results.
There is NO difference between having the indexes on the same disk or
different disk with the data while running pgbench. So I leave it up to
you guys as to include the patch or not. I do believe that even if
performance doesn't increase, this patch as alot of other benefits for
admins.

I bet it is the SCSI tagged queueing that is making up for the same disk
performance. Agreed administration is enough of a need to add the
feature.

-- 
  Bruce Momjian                        |  http://candle.pha.pa.us
  pgman@candle.pha.pa.us               |  (610) 853-3000
  +  If your life is a hard drive,     |  830 Blythe Avenue
  +  Christ can be your backup.        |  Drexel Hill, Pennsylvania 19026