You were brought to this page based on an internet search
and as a free service to Oracle DBAs.
The text below is an except from the book,
Oracle Performance Firefighting, written by
Craig Shallahamer of
OraPub, Inc.
Figures and tables are not included on this page, only their reference.
To order the book in either print or PDF form, click
here.
©2009, 2010 by Craig Shallahamer. This is copyrighted material.
PleaseOut of respect for those involved in the creation of the book and also for
their familes, we ask you to respect the copyright both in intent and deed. Thank you.
-------------------------------
* Find and reduce DML SQL impact. Since a free buffer waits event is related to too many dirty buffers in the LRU chain, there must be DML SQL. Depending on your performance tools, DML SQL can be difficult to find because it may be a high physical IO, logical IO, execution rate, or CPU consumer. It could be a sneaky combination of many statistics. If I cannot view the type of SQL (v$sql.c_type), then I look at both the top physical IO and logical IO SQL, and then check the statement itself. There is a good chance the DML SQL is also the top physical IO SQL statement as well (I see this a lot). If so, you've hit the jackpot and know you have found the key SQL statement. For more details about identifying DML SQL, refer to Chapter 8 section Application-Focused Solutions for Log File Parallel Write Contention. The objective is to reduce the impact, so that can mean tuning and also workload balancing, which will reduce the execution rate during peak times.
Buffers cannot be locked, because they are not relational structures. However, for a number of reasons, they can be temporarily unavailable. When this occurs, the buffer is deemed busy. Distilling all the complexity down to its essence, a buffer busy waits event is about limited concurrency. A process needs to access a buffer, but it can't because another process is accessing the buffer and won't allow concurrent access. There are a number of situations when this can occur and specific solutions in each case.
What makes buffer busy waits events challenging is there are multiple solid diagnostic approaches. One of the most common diagnostic buffer busy approaches is using what is known as the reason code. The reason code consists of three digits, with each digit's value revealing part of the story as to why the buffer is busy and cannot be immediately accessed.7 While many performance specialists use this method, I was never comfortable with it. Strangely, in Oracle Database 11g, Oracle removed the reason code from the v$session_wait and v$session view's p3 column! If you are used to using the reason code, don't fret, because the method I will detail works wonderfully.
©2009, 2010 by Craig Shallahamer. This is copyrighted material.
PleaseOut of respect for those involved in the creation of the book and also for
their familes, we ask you to respect the copyright both in intent and deed. Thank you.
|