Performance tuning without accessing real data [message #148482] |
Thu, 24 November 2005 07:19 |
rrawasi
Messages: 19 Registered: June 2004
|
Junior Member |
|
|
Dear All
My Boss want me to tune Oracle database, at first vision I see lots of problem in design.
the other bad thing is that they don't want to change the design and want to tune it.
So I can't gather lots of information about database, do you have any suggestion that Where should I invest on tuning database.
|
|
|
|
|
|
Re: Performance tuning without accessing real data [message #148576 is a reply to message #148534] |
Fri, 25 November 2005 02:22 |
rrawasi
Messages: 19 Registered: June 2004
|
Junior Member |
|
|
Thank you for your answer
I know How to tune my database very well but the problem is to don't touch the prodcution database, my boss don't trust me and the database is very very critical (T-com database)
I just think about some kind of replication that if it is possible have an exact same database (not just data within activity also) and focus tuning there.
OCP 9i, SCJP 1.4
[Updated on: Fri, 25 November 2005 02:33] Report message to a moderator
|
|
|
|
|
|
|
|
|
Re: Performance tuning without accessing real data [message #149205 is a reply to message #148482] |
Wed, 30 November 2005 09:35 |
smartin
Messages: 1803 Registered: March 2005 Location: Jacksonville, Florida
|
Senior Member |
|
|
One other option is to collect statistics from production and then import into your test environment. Not the data, just the statistics, which the test environment CBO would then use and be fooled into thinking that was the content of the data. Make sense?
Export the statistics from dictionary into your table. Export that table. Import that table export into your test database. Then import the statistics into the dictionary from that table. The dbms_stats package has procedures to do that. I think also Tom Kyte has talked about the procedure on his site.
It will give you an idea of the execution plans used. And in newer versions I think you can even export and import things like ADM statistics, all without touching the actual db data.
|
|
|
|
|
Re: Performance tuning without accessing real data [message #149401 is a reply to message #148482] |
Thu, 01 December 2005 08:43 |
smartin
Messages: 1803 Registered: March 2005 Location: Jacksonville, Florida
|
Senior Member |
|
|
Yes Frank that is a good point and a good link to read through, which I obviously hadn't done recently enough. I would view the technique as a last resort, but yet still incomplete and not always but only sometimes helpful approach. Using production data, or at the very least fake data that is simulated to look and act and have the same volume as production data is the best way to go. Simulating production data would be second best to using real production data, although the simulation process itself can be time consuming and labor intensive in some situations.
|
|
|
Re: Performance tuning without accessing real data [message #149490 is a reply to message #149401] |
Fri, 02 December 2005 00:56 |
Frank
Messages: 7901 Registered: March 2000
|
Senior Member |
|
|
I think you (Reza) will not be the first nor last DBA that is restricted from accessing the data there. To allow for performance-tuning and several other issues, access to the data is needed.
To being able to solve these issues, work on a way to 'anonymize' the data, rendering fake data.
This fake data should resemble the original in terms of histograms and index-skewness.
No short-term solution, I know, but in the long run it could help
hth
|
|
|
|