Oracle FAQ Your Portal to the Oracle Knowledge Grid
HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US
 

Home -> Community -> Mailing Lists -> Oracle-L -> RE: db file sequential read

RE: db file sequential read

From: Mark W. Farnham <mwf_at_rsiz.com>
Date: Mon, 21 Jun 2004 13:24:57 -0400
Message-ID: <KNEIIDHFLNJDHOOCFCDKMEKIEOAA.mwf@rsiz.com>


if the waits you are reporting are for physical file reads,

and if you have constructed your disk farm such that it has "stripe sets" that make up the files
underlying the tablespaces

and if you have at least three "stripe sets" that are independent in operation,

then put the table in a tablespace constructed on a file on "stripe set 1",

     put the primary index ..... on "stripe set 2",  and put the unique index ..... on "stripe set 3"

Then measure and see if it is faster.

Before you embark no this, make sure the service times you're seeing as the bottleneck are something
on the order of your physical disk speed, not the cache speed, since you're unlikely to get faster
than disk cache speed with separate physical threads, and you're not overruning cache if you're getting
service times faster than the physical disk speed.

If it does get faster, then your bottleneck actually was physical disk access, and you may or may not
have opportunities for additional speed up by constructing "stripe sets" on any remaining bottleneck
on faster media, or fine tuning such things I've left out of this somewhat simple minded layout as
where the rest of the indexes are, where the redo logs are, where undo is, where control files are, and
other competitive load on your system.

Further, in the construction of "stripe sets" folks far too often start each volume on the same disk.
To put it simply, if you have 8 drives, mirrored, so logically you have 4 drives, and you carve them
up into 4 volumes, start 1 on disk 1, 1 on disk 2, 1 on disk 3, and 1 on disk 4.

All of this assumes you've got physical file read waits to get your db file scattered read times.

mwf

-----Original Message-----

From: oracle-l-bounce_at_freelists.org
[mailto:oracle-l-bounce_at_freelists.org]On Behalf Of DENNIS WILLIAMS Sent: Monday, June 21, 2004 12:38 PM
To: 'oracle-l_at_freelists.org'
Subject: RE: db file sequential read

Harvinder

    Cache the index? Since Oracle must read the index to verify the uniqueness constraint you've created, and then to insert the new values in the indexes, I'm out of ideas here.

Dennis Williams
DBA
Lifetouch, Inc.
dwilliams_at_lifetouch.com

-----Original Message-----

From: oracle-l-bounce_at_freelists.org
[mailto:oracle-l-bounce_at_freelists.org]On Behalf Of Harvinder Singh Sent: Monday, June 21, 2004 11:37 AM
To: oracle-l_at_freelists.org
Subject: RE: db file sequential read

Dennis,

This option is not possible as we are just prototyping a particular part of application. When insers are going on other sessions can read data from this table and require indexes..we just need to tune inserts with indexes

Thanks
--Harvinder

-----Original Message-----

From: oracle-l-bounce_at_freelists.org
[mailto:oracle-l-bounce_at_freelists.org] On Behalf Of DENNIS WILLIAMS Sent: Monday, June 21, 2004 12:31 PM
To: 'oracle-l_at_freelists.org'
Subject: RE: db file sequential read

Harvinder

   What about dropping the indexes before inserting and recreating them afterward? This is a standard technique for increasing insert performance.

Dennis Williams
DBA
Lifetouch, Inc.
dwilliams_at_lifetouch.com=20

-----Original Message-----

From: oracle-l-bounce_at_freelists.org
[mailto:oracle-l-bounce_at_freelists.org]On Behalf Of Harvinder Singh Sent: Monday, June 21, 2004 9:45 AM
To: oracle-l_at_freelists.org
Subject: db file sequential read

Hi,

During testing the insert performance in oracle 10g on win2k, we see the wait event very high for the db file sequential read. With trace 10046,8 I can see that all the blocks are in file that contains primary and unique index for the table. How can we reduce the above waits on these indexes. Do freelist and initrans parameter still valid for 10g and what should be the reasonable value. We are inserting 1M records 1000 at a time by selecting from other table. Also if I try to insert from 2 session simultaneously insert become more slow.

Thanks
--Harvinder

=3D20



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

-----------------------------------------------------------------
----------------------------------------------------------------
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
Received on Mon Jun 21 2004 - 12:27:57 CDT

Original text of this message

HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US