So, the agenda for my presentation at the UKOUG BIRT SIG on January 31st is going to be something like this:
Abstract: Just like OLTP systems, Data Warehouse performance can be affected by a myriad of factors. This presentation will cover Five key areas which can have a significant impact on warehouse performance:
- Make the most of your memory
- Partition for performance & manageability
- Squeeze your data with segment compression
- Beware Of Temporal Data Affecting The Optimizer
- Find Out Where Your Query Is At
What other first aid tips do you think should be mentioned? Leave us a reply and let us know! Importance viagra for cheap prices of pet health supplements Good nutrients are not enough for your pet’s good health. The medication is not intended for use by women or by viagra levitra cialis children. Due to its numerous benefits, Chiropractic therapy can have diverse http://greyandgrey.com/wp-content/uploads/2018/07/Jansch.pdf best viagra pills applications both in the world of sports and exercise-related injuries. Those who suffer from the following cost of cialis should consult their doctor before using this medicine.
Each of these are major topics in their own right so I’m not going to cover an enormous amount of ground with each, but I will try to cover some key points and include a demonstration or two along the way.
I’ll be putting the slides together over the next few weeks so I guess I’ll be posting less (already have been…but that’s due to this man flu still hanging around).
I have come across a couple of interesting things along the way already mind:
Firstly, I was trying to tune a Materialized View the other day when I encountered something which made me laugh. I extracted the DDL from Toad and tried to do an explain plan on it…only to discover that it rewrote the query using the Materialized View – well, of course, it would, wouldn’t it! In order to tune it you need to add the NOREWRITE hint.
My colleague Anthony Evans encountered a similar problem afterwards when trying to tune a query which referred to a view which was undergoing a Materialized View Rewrite when he didn’t want it to. The solution he found in this case was to use the NO_QUERY_TRANSFORMATION hint to instruct the optimizer to skip all query transformations.
The next thing I discovered was when building a test script to test compression out. I was doing a series of INSERT /*+ APPEND */ operations with COMMITs after them which was working fine. Then I added SET AUTOTRACE ON at the start of the script so I could start to see the execution plans and statistics but unfortunately this resulted in this error:
ORA-12838: cannot read/modify an object after modifying it in parallel
The AUTOTRACE is doing some more DML after the INSERT /*+ APPEND */ but before the COMMIT…which is not allowed when using the APPEND and so it breaks.
I resorted to using SET AUTOTRACE ON STATISTICS instead since I wasn’t too interested in the execution plan anyway. I guess I could have used tracing and tkprof if I’d wanted to see the execution plans as well.
