Re: Write Complete Waits

Hi Neil,

Yes, the internal write batch size is derived from these parameters. The basic formula is db_files * db_file_simultaneous_writes / 2. There are some other limits such as db_block_buffers and the platform specific maximum number of concurrent asynchronous I/O operations, but those would not be relevant in this case.

Regards,
Steve Adams

http://www.ixora.com.au/

http://www.oreilly.com/catalog/orinternals/

http://www.christianity.com.au/

On Thu, 9 Dec 1999 14:36:19 -0000, "NeilC" <plums_at_bigfoot.com> wrote:

>Steve,
>
>What is the effect of reducing db_files and db_file_simultaneous_writes ,
>presumably these are used in an internal algorithm as part of a calculation
>of some sort ?
>
>Regards
>
>Neil
>
>Steve Adams wrote in message <384eec92.38295185_at_news.eagles.bbs.net.au>...
>>Hi Roy,
>>
>>In general, 'write complete waits' indicate either that the internal
>>write batch size is too large, or that you are checkpointing hot
>>blocks too intensively, rather than that write I/O is too slow. With
>>only three physical disks, you should set db_files as low as possible
>>and set db_file_simultaneous_writes to 1. Then set
>>db_block_checkpoint_batch to db_files/4.
>>
>>If you are using incremental checkpoints (db_block_max_dirty_target),
>>don't be too ambitious. Otherwise make sure that background
>>checkpoints are not too frequent.
>>
>>Hope this helps,
>>Regards,
>>Steve Adams
>>http://www.ixora.com.au/
>>http://www.oreilly.com/catalog/orinternals/
>>http://www.christianity.com.au/
>>
>>
>>On Wed, 08 Dec 1999 21:51:39 GMT, rspeaker_at_my-deja.com wrote:
>>
>>>I currently have an Oracle 8.0.4 system running on an RS/6000 with 4x332
>>>mhz CPU and 1 GB RAM. I have async I/O configured on the system. The
>>>application in question has its 'data' file striped across 2 disks, and
>>>the corresponding 'index' file on 1 disk. I have been seeing as much as
>>>10% write complete waits throughout the day (according to the monitoring
>>>tool I am using). I understand the fix for this to be either enabling
>>>Async I/O or starting multiple db writers with the dbwr_io_slave
>>>parameter. I have async I/O enabled but still see the waits. Will it
>>>help/hurt performance to add a second db writer as well? Any other
>>>suggestions?
>>>
>>>thanks,
>>>Roy
>>>
>>>
>>>Sent via Deja.com http://www.deja.com/
>>>Before you buy.
>>
>
>
Received on Thu Dec 09 1999 - 15:21:20 CST