Re: VxFS threaded AIO vs. DBWR Slaves

From: Charlotte Hammond <>
Date: Tue, 26 Feb 2008 13:24:45 -0800 (PST)
Message-ID: <>

Hi Greg,

Thanks for the tip. Unfortunately I'm not in a position to dictate ASM at present and have to live with what I've got (at least for now!).

Somebody mentioned off-list that threaded AIO may not be a feature of VxFS on HP-UX (although I believe it is on Solaris). Can anyone confirm or refute? I'm finding it hard to get a definitive answer on this.

Assuming that threaded AIO *does* exist and that I *do* want to use it, should I just set FILESYSTEMIO_OPTIONS=SETALL and DISK_ASYNCH_IO=TRUE as if it were kernelized AIO?

Thanks You!

  • Original Message ---- From: Greg Rahn <> To: Cc: Sent: Monday, February 25, 2008 10:09:29 PM Subject: Re: VxFS threaded AIO vs. DBWR Slaves

I would suggest using ASM. Then you can leverage all the benefits and not have to worry about file layout.

If you use threaded async or emulate async by multiple db writers, its probably the same overhead.

On 2/25/08, Charlotte Hammond <> wrote:
> Hi All,
> I'm building a 10g database on a VxFS filesystem on HP-UX. I understand this does not support kernelized asynchronous I/O but does support threaded asynchronous I/O which I believe has a significant CPU overhead. So: should I use the threaded asychronous I/O from VxFS (FILESYSTEMIO_OPTIONS=SETALL) or should I switch it off, and use multiple database writer slaves instead?
> My aim is targeted towards performance rather than capacity (i.e. small number of users in database).


Greg Rahn

Looking for last minute shopping deals?  
Find them fast with Yahoo! Search.
Received on Tue Feb 26 2008 - 15:24:45 CST

Original text of this message