Re: direct read is much faster than scattered read, is it reasonable?

  • From: "Alexander Fatkulin" <afatkulin@xxxxxxxxx>
  • To: "Zhu,Chao" <zhuchao@xxxxxxxxx>
  • Date: Wed, 28 May 2008 16:23:22 -0400

Did you mean that operations like FTS/IFFS can result in "scattered read" as
well as "direct path read"? That is true and stats will be logged against
relative operation.

On Wed, May 28, 2008 at 3:52 PM, Zhu,Chao <zhuchao@xxxxxxxxx> wrote:

> some direct path read can caused by "scattered read" as well, like parallel
> read into pga, and when "serial_direct_path_read=true"
>
> On Thu, May 29, 2008 at 1:08 AM, Alexander Fatkulin <afatkulin@xxxxxxxxx>
> wrote:
>
>> You can take a look at v$event_histogram to see how distribution goes
>>
>> don't forget that direct path reads are done by 1M chunks
>> (_db_file_direct_io_count) while scattered read are driven by
>> db_file_multiblock_read_count -- it might as well be that your SAN is able
>> to do better prefetching in case of 1M chunks, you can find out by altering
>> dbfmrc at the session level to a different values and see how it will affect
>> the results
>>
>>>
>>>
>>> Thanks,
>>> Qihua
>>>
>>
>>
>>
>> --
>> Alex Fatkulin,
>> The Pythian Group,
>> http://www.pythian.com/blogs/author/alexf
>>
>
>
>
> --
> Regards
> Zhu Chao
> www.cnoug.org
>



-- 
Alex Fatkulin,
The Pythian Group,
http://www.pythian.com/blogs/author/alexf

Other related posts: