Posted by Daniel Fink on Dec 1, 2010
My presentations are finished and the feedback positive (“This was not what I had expected, but it is really making me think.”). I will be presenting one of sessions (Using AWR/ASH data) at RMOUG Training Days and Hotsos Symposium 2011 with minor modifications (I wonder…can I use the 11g analytical function LISTAGG with AWR/ASH data?).
Highlights from the last day and a half…
- AWR Time Series Analysis (Breitling) – Don’t underestimate the benefit of vizualization in analysis.
- Co-operating with the database (Lewis) – Bad systems fail because they are bad and nobody uses them; Good systems fail because they are good and everybody uses them.
- ASM Failability (Gorbachev) – There is no configuration that protects you 100% of the time.
- Advanced Oracle Troubleshooting, Part 1 (Geist) – Finding a good tuning fix without any side effects is not a trivial task.
- ASM Redundancy Levels (Hawker) – Automated processes and disaster recovery can have unintended, unforeseen and disasterous side effects.
- Back to the Basics (Poder) – Performance diagnosis and optimization is an art…only to those who do not understand fundamentals.
- Advanced Oracle Troubleshooting, Part 2 (Geist) – Cursor dumps will show the bind variable value used to parse the plan and the current value for the execution.
Now…in my own defense…my blog posts are short because I value my reader’s time and want to deliver high informational throughput!
Well…the reality is that I left my netbook’s power supply at home and my battery power is limited to only essential and brief work. I shall endeavor to expound on my adventures under regular power later in the week!
Posted by Daniel Fink on Nov 30, 2010
After an inauspicious start (note to self – create a travel packing checklist), my 7th visit to Birmingham for the UKOUG Conference is in full swing. As in the past, the quality of the presentations is outstanding and the biggest challenge to the attendee is to decide which of the 3 concurrent sessions to attend.
- Large Scale ASM Adoptions and Lessons Learned (Jeremy Schneider) – #1 success factor for ASM implementations is the buy in from Systems Teams.
- Mostly Hidden Mostly Free Oracle Features (Graham Wood) – If you are using AWR and distributing the report, always include an ADDM report. It may not be 100% correct, but it does provide an automatic first level analysis.
- Practical ASH (David Kurtz) – For systems with lots of literal sql, different statements with the same plan hash value are often identical, except for the literal values.
- Top Tips for the CBO (Maria Colgan) – Cardinality misestimation is the top problem for the optimizer.
- Time Series Analysis (Wolfgang Breitling) – Combining data and statistical techniques can help identify specific changes that impact performance and predict future performance.
Posted by Daniel Fink on Sep 13, 2010
In preparing my presentation for the Michigan Oak Table Symposium, I came across AWR extract and load. While these statements are documented in the Oracle manuals (kind of), I don’t see much discussion online, which is a good barometer for the popularity of an item. Not much discussion – not very well known or used. Although the scripts are in $ORACLE_HOME/rdbms/admin in 10.2, they are not documented.
One of the frustrations with AWR (and Statspack) has been that diagnosing past performance issues and trending are dependent on keeping a large number of snapshots online. This means more production storage, resource consumption with queries. Would it not be nice to be able to take the AWR data from your production system, load it into a dba repository and then do all your querying? Perhaps even your own custom ETL to pre-aggregate, create custom views and procedures?
As of 10.2, Oracle supplies two scripts that enable you to extract and load AWR data into another database (even one already running AWR snapshots). You can even take the AWR data from a 10.2 database on Solaris and load it into an 11.2 database on Windows XP (other variations may work…but these are the two versions I have handy). I also took 11.2 database on Linux and loaded it to the Windows database.
Read the rest of this entry . . .
