With a functional Oracle Disk Manager Library
linked in, this is a non issue.
do ldd $ORACLE_HOME/bin/oracle .. see one of the
first libs is a "stub" library called libodm9.so which
is a symlink to libodm9d.so
if you have a functional libodm9.so (such as that in the
VRTS foundation suite or Polyerve Matrix Server on Linux),
Oracle ceases to make calls for IO through the traditional
OSD routines skfg()* ..instead, all IO is odm_io() which
is async...
you can read more about this in the in the whitepaper
section on www.polyserve.com
-----Original Message-----
From: McAllister, Andrew [mailto:McAllisterA@(protected)]
Sent: Wednesday, August 18, 2004 8:18 AM
To: suse-oracle@(protected)
Subject: RE: [suse-oracle] SLES 9 Async IO broken with Oracle 9i
Indeed. Many people don't realize that async io is so useful.
It basically "parallelizes" IO tasks. Obviously, in a multi
user environment running 10 parallel IO tasks will be MUCH
faster than 10 sequential ones.
Unfortunately, async io doesn't appear to be very mature on
Linux. For example the "current" aio routines are built-in to
the 2.6 kernel. But those interfaces are apparently different
from those in SLES 8. None of the current aio "testing"
programs I found will even compile on SLES 8.
Also, Suse appears to be breaking AIO (with Oracle) even in
SLES 8 (as of -226 and above). We can only run kernel
2.4.21-203 or 2.4.21-215 when we have AIO turned on. Otherwise
our DB crashes and won't restart without an OS reboot.
Like you, I really hope SuSE and Oracle get serious about this
issue. I've got $20,000 boxes that run SLES 8 that run circles
around practically any other piece of gear we've got (and we
have some VERY expensive gear). But, convincing others to put
up with the quirks of Oracle/Linux is getting harder to do.
SuSE help us all!
> -----Original Message-----
> From: Anton Dischner [mailto:Anton.Dischner@(protected)]
snip
> we have a server with single fibre channel.
> The performance dropped significantly with users > 1.
>
> Switching on async io solved this problem.
>
> No async IO in SLES9 would be a severe show stopper for this box.
>
> I hope SuSE and Oracle get serious about this issue.
snip
--
To unsubscribe, email: suse-oracle-unsubscribe@(protected)
For additional commands, email: suse-oracle-help@(protected)
Please see http://www.suse.com/oracle/ before posting
--
To unsubscribe, email: suse-oracle-unsubscribe@(protected)
For additional commands, email: suse-oracle-help@(protected)
Please see http://www.suse.com/oracle/ before posting