RE: Performance Question - High I/O per Insert
- From: "Mark Strickland" <mstrickland@xxxxxxxxxxxxx>
- To: <Graeme.Farmer@xxxxxxxxxx>
- Date: Tue, 12 Oct 2004 16:50:01 -0700
Thanks, I'll check that out!
________________________________
From: Graeme.Farmer@xxxxxxxxxx [mailto:Graeme.Farmer@xxxxxxxxxx]
Sent: Tuesday, October 12, 2004 4:38 PM
To: Mark Strickland
Cc: oracle-l@xxxxxxxxxxxxx; oracle-l-bounce@xxxxxxxxxxxxx
Subject: RE: Performance Question - High I/O per Insert
Have you tracked the "branch node splits" and "leaf node splits"
statistics from v$sesstat for the session(s) performing the inserts?
If you have STATSPACK collecting information you should see an upward
trend for these stats towards the end of the week but it would be best
to target the inserting sessions to be more conclusive in testing your
hypothesis. A logoff trigger may be necessary if the sessions are
transient.
Graeme
"Mark Strickland" <mstrickland@xxxxxxxxxxxxx>
12/10/2004 02:41 AM
To
<Graeme.Farmer@xxxxxxxxxx>
cc
<oracle-l@xxxxxxxxxxxxx>, <oracle-l-bounce@xxxxxxxxxxxxx>
Subject
RE: Performance Question - High I/O per Insert
No, no triggers. I believe the high I/O is due to index maintenance.
When I remove the indexes or run the load with direct path, the load
screams. Thx!
Mark
________________________________
From: Graeme.Farmer@xxxxxxxxxx [mailto:Graeme.Farmer@xxxxxxxxxx]
Sent: Sunday, October 10, 2004 5:52 AM
To: Mark Strickland
Cc: oracle-l@xxxxxxxxxxxxx; oracle-l-bounce@xxxxxxxxxxxxx
Subject: RE: Performance Question - High I/O per Insert
Any triggers on the table that refer to data found in the current week's
partition? This is a classic cause of high I/O per insert.
Graeme
oracle-l-bounce@xxxxxxxxxxxxx wrote on 29/09/2004 05:27:28 AM:
> This is a follow-up to a question I posed earlier, but I have more
info.
> A SQL*Loader conventional path load inserts rows into a table
> partitioned by week. On the first day of a new week, the load runs as
> efficiently as expected, but as the week goes on, the load goes more
and
> more slowly. There is a single local index on each partition and the
> blevel is 2. I'm scratching my head. I realize I'm not providing
much
> info, but can anyone tell me why loading into a partition might get
> slower as the partition fills? Thx!
>
>
> Mark Strickland
>
>
>
> ________________________________
>
> From: Mark Strickland
> Sent: Wednesday, September 22, 2004 12:42 PM
> To: oracle-l@xxxxxxxxxxxxx
> Subject: Performance Question - High I/O per Insert
>
>
>
> I'm trying to understand why I'm seeing a high number of I/Os for
> inserts. Inserts are through SQL*Loader conventional patch. Average
> row length is 129 bytes, no longs or blobs. The table has five
indexes
> and each index has a blevel of 2. According to v$sqlarea, each insert
> uses 650 logical I/Os, 69 physical I/Os. I would expect fewer than 20
> I/Os per insert. There are db file sequential read waits on the data
> files that make up the index tablespace. That file system also
contains
> the archived logs. Not surprised at the contention. Can someone
point
> me in the right direction to understand this? Thx.
>
>
>
> Mark Strickland
>
> Drugstore.com
>
> Seattle, WA
>
>
> --
> //www.freelists.org/webpage/oracle-l
--
This transmission is for the intended addressee only and is confidential
information. If you have received this transmission in error, please
notify the sender and delete the transmission. The contents of this
e-mail are the opinion of the writer only and are not endorsed by the
Mincom Group of companies unless expressly stated otherwise.
--
This transmission is for the intended addressee only and is confidential
information. If you have received this transmission in error, please
notify the sender and delete the transmission. The contents of this
e-mail are the opinion of the writer only and are not endorsed by the
Mincom Group of companies unless expressly stated otherwise.
--
//www.freelists.org/webpage/oracle-l
Other related posts:
