RE: Solaris 8 and asynch IO

In this thread raw and Quick IO are not an option.

I have read the papers on ixora that touch on the subject, but was wanting to hear from others. I have read most of what is out there on the subject although I dont claim to remember it all.

--
- David / Thump
Life is what happens while waiting 
or planning for the future.


> 1.)  If you can do kaio, (raw or VxFS+QIO), set disk_async_io=TRUE and be happy.

> 2.)  If you can't do kaio, set disk_async_io=FALSE.  As to slaves vs. multiple 

> writers, it depends on your environment and the type of load you're supporting.  

> I'll just direct you to Steve Adams' site, http://www.ixora.com.au/ for more 

> details.  Do a search on 'dwbr io slave', and you'll find lots to read.

> 3.)  If you have a combination of raw and filesystem files, you should also 

> investigate the 'filesystemio_options' parameter.  Again, a good explanation is 

> available on Steve's site.

> 

> 

> 

> -----Original Message-----

> From: thump604_at_comcast.net [mailto:thump604_at_comcast.net]

> Sent: Friday, March 19, 2004 2:40 PM

> To: oracle-l_at_freelists.org

> Subject: Solaris 8 and asynch IO

> 

> 

> Happy Friday everyone!

> 

> I would like to hear from some others regarding their opinions and feedback in 

> terms of asynch IO on Solaris when Veritas is not in the mix.

> OS Solaris 8

> DB 8174

> 

> We know that if Quick IO or raw devices are not available that Solaris only 

> simulates asynch IO with a failed kaio call and then the use of light weight 

> threaded processes if disk_asycnh_io is set to true.  If disk_asynch_io is set 

> to false and dbwr_io_slaves are employed, then asycnh IO is simulated, but this 

> time at the database level.  My understanding is that IO slaves in terms of 

> simulating asycnh IO can setrvice IO requests for most of the Oracle internal 

> processes, where as OS pwrite/pread calls are more limited in what requests they 

> can service.

> 

> With that, the question remains is there a hard answer or actual best practice 

> in terms of disk_asycnch_io=false/true combined with db_writer_processes VS 

> dwr_io_slaves?

> 

> I believe that the setting of false and 4 dbwr_io_slaves is optimal.

> I would like to hear what others think or have found on this matter.

> --

> - David / Thump

> Life is what happens while waiting 

> or planning for the future.

> ----------------------------------------------------------------

> Please see the official ORACLE-L FAQ: http://www.orafaq.com

> ----------------------------------------------------------------

> To unsubscribe send email to:  oracle-l-request_at_freelists.org

> put 'unsubscribe' in the subject line.

> --

> Archives are at http://www.freelists.org/archives/oracle-l/

> FAQ is at http://www.freelists.org/help/fom-serve/cache/1.html

> -----------------------------------------------------------------

> ----------------------------------------------------------------

> Please see the official ORACLE-L FAQ: http://www.orafaq.com

> ----------------------------------------------------------------

> To unsubscribe send email to:  oracle-l-request_at_freelists.org

> put 'unsubscribe' in the subject line.

> --

> Archives are at http://www.freelists.org/archives/oracle-l/

> FAQ is at http://www.freelists.org/help/fom-serve/cache/1.html

> -----------------------------------------------------------------

----------------------------------------------------------------
Please see the official ORACLE-L FAQ: http://www.orafaq.com
----------------------------------------------------------------
To unsubscribe send email to:  oracle-l-request_at_freelists.org
put 'unsubscribe' in the subject line.
--
Archives are at http://www.freelists.org/archives/oracle-l/
FAQ is at http://www.freelists.org/help/fom-serve/cache/1.html
-----------------------------------------------------------------