From: Josh Berkus <josh@agliodbs.com>
Organization: PostgreSQL @ Sun
To: pgsql-performance@postgresql.org
Subject: Need help with 8.4 Performance Testing
Date: Sun, 7 Dec 2008 11:38:01 -0800
User-Agent: KMail/1.8.2
Mime-Version: 1.0
Content-Type: text/plain
Content-Transfer-Encoding: quoted-printable
Message-Id: <200812071138.01215.josh@agliodbs.com>

Database performance geeks,

We have a number of patches pending for 8.4 designed to improve database=20
performance in a variety of circumstances.  We need as many users as =
possible=20
to build test versions of PostgreSQL with these patches, and test how =
well=20
they perform, and report back in some detail.

Particularly, users with unusual hardware architectures (16 or more =
cores,=20
ARM, Power, SSD, NFS-mounted data) or operating systems (Solaris, OSX,=20
Windows-64) are really helpful.  Testers need to be familiar with =
building=20
PostgreSQL from source and patching it, as well as basic PostgreSQL =
Tuning=20
(except for the Wizard Patch) and have some kind of performance test=20
available, ideally something based on your own application use.

If you are going to use pgbench to test, *please* read Greg Smith's =
notes=20
first:=20
http://www.westnet.com/~gsmith/gregsmith/content/postgresql/pgbench-scali=
ng.htm

The Wiki (http://wiki.postgresql.org/wiki/CommitFest_2008-11) has a full =
list=20
of patches, but below are the ones in particular we could use help with.

You *do* need to read the entire mail threads which I link to below to=20
understand the patches.  Thanks for your help!

Proposal of PITR performance improvement (Koichi Suzuki):
http://archives.postgresql.org/message-id/a778a7260811270404g49254640x8ed=
58b12b7c65d0b@mail.gmail.com
http://archives.postgresql.org/message-id/a778a7260810280033n43f70d36x8c4=
37eacf9a5461e@mail.gmail.com

Simple postgresql.conf wizard
http://archives.postgresql.org/message-id/Pine.GSO.4.64.0811012101220.176=
19@westnet.com
http://archives.postgresql.org/message-id/Pine.GSO.4.64.0811291403040.128=
85@westnet.com

Improve Performance of Multi-Batch Hash Join for Skewed Data Sets
http://archives.postgresql.org/message-id/6EEA43D22289484890D119821101B1D=
F2C1683@exchange20.mercury.ad.ubc.ca
http://archives.postgresql.org/message-id/1924d1180811051606w19aaf30du589=
e8ea10ea5534d@mail.gmail.com
http://archives.postgresql.org/message-id/22901.1227228246@sss.pgh.pa.us

Window Functions
http://archives.postgresql.org/message-id/e08cc0400810270912u49a6ec83vc23=
984c01f368f76@mail.gmail.com
http://archives.postgresql.org/message-id/492D1246.5070101@enterprisedb.c=
om
http://archives.postgresql.org/message-id/839FB90FF49D4120B7107ED0D7B3E5B=
6@amd64

parallel restore
(especially need to test on 16+ cores)
http://archives.postgresql.org/message-id/490878AC.1@dunslane.net

B-Tree emulation for GIN
http://archives.postgresql.org/message-id/491B1888.9020903@sigaev.ru
http://archives.postgresql.org/message-id/Pine.LNX.4.64.0811191828050.786=
2@sn.sai.msu.ru

Also, the following patches currently still have bugs, but when the bugs =
are=20
fixed I'll be looking for performance testers, so please either watch =
the=20
wiki or watch this space:

=2D- Block-level CRC checks (Alvaro Herrera)
=2D- Auto Partitioning Patch (Nikhil Sontakke)
=2D- posix_fadvise (Gregory Stark)
=2D- Hash Join-Filter Pruning using Bloom Filters
=2D- On-disk bitmap indexes

Please report your results, with the patchname in the subject line, on =
this=20
mailing list or on -hackers.  Thank you, and your help will get a better =
8.4=20
out sooner.

=2D-=20
Josh Berkus
PostgreSQL
San Francisco