Re: Filesystem device O/S Buffering

Oracle does in fact use the O_SYNC mode of the open() api call for all writes to the filesystem. This ensures that data is actually written through the OS buffer cache when Oracle writes to any file. However this is not true for Hardware cache etc.

In article <74ou6b$90l$1_at_sunrise.pg.gda.pl>,   "Piotr Kolodziej" <pkol_at_otago.gda.pl> wrote:
> Art S. Kagel wrote in message <366FF90A.6EB2_at_bloomberg.net>...
> >The fflush() function only flushes the application level buffers in the =
>
> >stdio library to be flushed to the filesystem's buffers it does not=20
> >flush the filesytem buffers to disk. That is ONLY accomplished by=20
> >opening the file O_SYNC which causes all I/O to the file to be flushed=20
> >from the buffer cache to disk immediately.
>
> You're right. It is a very good explanation how Oracle may ommit
> OS buffering and ensure flushing of redo during commit.
>
> Piotr Kolodziej
> pkol_at_otago.gda.pl
>
>

-----------== Posted via Deja News, The Discussion Network ==---------- http://www.dejanews.com/ Search, Read, Discuss, or Start Your Own Received on Mon Dec 14 1998 - 12:38:17 CST