Re: Async I/O with Oracle 7.1.3 or 7.1.6 on HP/UX

From: Alan J. Heckman <aheckman_at_us.oracle.com>
Date: 1995/06/12
Message-ID: <3ri81v$je1_at_inet-nntp-gw-1.us.oracle.com>#1/1


Donald Campbell (Donald_at_suedon.demon.co.uk) wrote:
> In article <rp1399-0506951403470001_at_m12cim15.sps.mot.com>
> rp1399_at_email.sps.mot.com "Steve Lemme" writes:
 

> > > I am trying to find someone who has set up Async I/O on an
> > > HP/UP v9.04 with either Oracle 7.1.3 or 7.1.6.
> > >
> > > We have been talking to Oracle "Support", and they have
> > > said that we should set the ANSYC_WRITE=true in the INIT.ORA. The
> > > only problem is that the parameters does NOT exist (7.1.3). We pointed this
> > > out, and "Support" has said that the parameters will be supported in
> > > 7.1.6. I then talked to a friend who does have 7.1.6 installed, and
> > > guess what....no parameter.
> >
> >
> > We use it. It is enabled as an HPUX Kernel parameter.
> >
> >
> > Steve Lemme
> > DBA
> > MOTOROLA
> >
 

> Async writes have been available under Dynix/ptx for some time now. The
> primary functionality is in the Kernel. You enable its use by Oracle RDBMS
> by setting the flag in init.ora. As per the documentation it is for use
> with raw devices.
 

> Regards
> Donald Campbell [ SueDon Ltd - UNIX and Database Specialists ]

Async I/O implementation is pretty platform specific.

On HP/UX you configure the HP/UX kernel to include the async driver, set up the device file correctly, put all database files on raw devices, and use only one database writer. Async I/O will then be used automatically. Check out the latest dbwr* logs to see if it is working. There is no other init.ora parameter other than specifying only one db writer. Check out the IUG for Oracle on HP for detail on setting up the async driver. The "ANSYC_WRITE=true" is for some other ports.

There is no "kernel parameter" to set. If you set fs_async you are just telling the filesystem to use async I/O. Oracle opens the datafiles with O_SYNC so it won't affect anything although your other files on filesystems may be corrupted when the system crashes.

--
/====================================================================\
| Alan J. Heckman    PGPPrint 1024/30FB29CD  71 BA B2 E8 EF AA 65 3A |
| aheckman_at_us.oracle.com                     23 5D 1C 0E B8 79 70 03 |
\====================================================================/
Received on Mon Jun 12 1995 - 00:00:00 CEST

Original text of this message