Re: Commit -> write to data file immediately???

From: andrew_webby at hotmail <spam_at_no.thanks.com>
Date: Mon, 2 Apr 2001 12:14:37 +0100
Message-ID: <986210092.16064.0.nnrp-09.c30bdde2@news.demon.co.uk>

fwiw, I have a faint memory that there's a solaris/oracle related tuning parameter where timestamps are not updated on data files every time a write is made. Improves performance supposedly (well, it would make sense I suppose).

"Jonathan Lewis" <jonathan_at_jlcomp.demon.co.uk> wrote in message news:986053458.11410.1.nnrp-02.9e984b29_at_news.demon.co.uk...
>
> The timestamps on files are not necessarily relevant.
> HP-UX, for example, has (?had) two flags for writing
> to files OSYNC and ODSYNC. One updates the file,
> the other updates the file and the inode.
>
> At some level of HP-UX and Oracle, performance on
> writes bombed because the wrong flag was used by
> Oracle, and an o/s workaround was needed to bypass
> the problem (setting osync_is_odsync).
>
>
> --
> Jonathan Lewis
> Yet another Oracle-related web site: http://www.jlcomp.demon.co.uk
>
> Practical Oracle 8i: Building Efficient Databases
> Publishers: Addison-Wesley
>
> Reviews at: http://www.jlcomp.demon.co.uk/book_rev.html
> >
> > however, as Frank (also me) mentioned, the timestamp of all redo log
> >files remain the same, so LGRW writes nothing to the redo log files. How
> >(and where) Oracle restore the committed record? May be after pulling
> >the plug, and before the computer turn off, Oracle writes the entry to
> >the redo log....
> >
>
>
>
Received on Mon Apr 02 2001 - 06:14:37 CDT