Nice Oracle tuning article

This is a nice article about tuning Oracle for Linux:

http://www.linuxjournal.com//article.php?sid=5840

It would be nice to have some of these numbers for PostgreSQL. Does
no-access-time affect performance? What about the other parameters?

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

Bruce Momjian <pgman(at)candle(dot)pha(dot)pa(dot)us> writes:
> This is a nice article about tuning Oracle for Linux:
> http://www.linuxjournal.com//article.php?sid=5840

I thought it was a tad amusing that his benchmarking setup was TPC-C
driven by Benchmark Factory.

(For those not in the know, this is exactly the setup that Great Bridge
used a couple years ago for their much-maligned benchmarks... One would
at least like to know what sort of networking is involved, seeing as
how Benchmark Factory only runs on Windows.)

regards, tom lane

On Sun, 24 Feb 2002 20:42:15 -0500 (EST)
Bruce Momjian <pgman(at)candle(dot)pha(dot)pa(dot)us> wrote:

> This is a nice article about tuning Oracle for Linux:
>
> http://www.linuxjournal.com//article.php?sid=5840
>
> It would be nice to have some of these numbers for PostgreSQL. Does
> no-access-time affect performance? What about the other parameters?
Bruce,

I know for a fact that under FreeBSD setting noatime in /etc/fstab helps both
PostgreSQL and Apache. Otherwise the system has to update whichever file has been
touched even though on 95% of the files we don't care. Maybe not quite as a big win
as Apache, but it does help PostgreSQL.

GB

--
GB Clark II | Roaming FreeBSD Admin
gclarkii(at)VSServices(dot)COM | General Geek
CTHULU for President - Why choose the lesser of two evils?

Tom Lane wrote:
> Bruce Momjian <pgman(at)candle(dot)pha(dot)pa(dot)us> writes:
> > This is a nice article about tuning Oracle for Linux:
> > http://www.linuxjournal.com//article.php?sid=5840
>
> I thought it was a tad amusing that his benchmarking setup was TPC-C
> driven by Benchmark Factory.
>
> (For those not in the know, this is exactly the setup that Great Bridge
> used a couple years ago for their much-maligned benchmarks... One would
> at least like to know what sort of networking is involved, seeing as
> how Benchmark Factory only runs on Windows.)

Benchmark Factory connects via ODBC to a database and runs
some queries. The queries might look similar to what a TPC-C
benchmark would look like, but be sure, it has nothing to do
with a real TPC-C implementation.

I analyzed the thing during my time at GB a little. There is
no system under test, stressed by simulated users. Instead
there is a simulated application stressing a database. The
database schema uses floating point fields for monetary data,
making the run of the constistency checks impossible. The
consistency after a benchmark run wouldn't be given anyway
because it doesn't really do what TPC-C defines.

And most important, it doesn't allow you to use the strength
of your database system! The TPC-C specs for example allow
you to move part of the logic into stored procedures and
functions. Impossible with that Benchmark Factory stuff.

All in all it produces some nice graphics, but that's it
IMNSVHO.

Jan

--

#======================================================================#
# It's easier to get forgiveness for being wrong than for being right. #
# Let's break this rule - forgive me. #
#================================================== JanWieck(at)Yahoo(dot)com #

_________________________________________________________
Do You Yahoo!?
Get your free @yahoo.com address at http://mail.yahoo.com

GB Clark wrote:
> On Sun, 24 Feb 2002 20:42:15 -0500 (EST)
> Bruce Momjian <pgman(at)candle(dot)pha(dot)pa(dot)us> wrote:
>
> > This is a nice article about tuning Oracle for Linux:
> >
> > http://www.linuxjournal.com//article.php?sid=5840
> >
> > It would be nice to have some of these numbers for PostgreSQL. Does
> > no-access-time affect performance? What about the other parameters?
> Bruce,
>
> I know for a fact that under FreeBSD setting noatime in /etc/fstab helps both
> PostgreSQL and Apache. Otherwise the system has to update whichever file has been
> touched even though on 95% of the files we don't care. Maybe not quite as a big win
> as Apache, but it does help PostgreSQL.

I would be interested to know if this parameter setting gives the same
performance boost as noatime:

wal_sync_method = fdatasync

Does FreeBSD have fdatastync? Should be the same because atime
performance is only an issue with fsync, and fdatasync is supposed to
only update the data and not the inode stuff.

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