If you purchase Jonathan Lewis’ “Oracle Core” book, you will have access to a couple of the scripts that he wrote for capturing perfomance statistic deltas. I think that the two scripts that you are interested in are snap_myst.sql and snap_stat.sql
http://jonathanlewis.wordpress.com/oracle-core/

Tanel Poder offers a free utility called Snapper. I personally have not used it, but I have heard many favorable comments about what it does:
http://tech.e2sn.com/oracle-scripts-and-tools/session-snapper

My Hyper-Extended Oracle Performance Monitor program will also do that, you can see the screen captures here:
http://hoopercharles.wordpress.com/2012/03/12/thoughts-on-a-hyper-extended-oracle-performance-monitor-beta/

In one of the book chapters that Randolf Geist and I co-wrote, we included a simple script that calculates the delta values of various performance views – this might be a good starting point:
http://books.google.com/books?id=tUNtyMCwDWQC&pg=PA203

On page 212 in the same book, we showed how to do the same with Statspack snapshots (with a timed delay between captures):
http://books.google.com/books?id=tUNtyMCwDWQC&pg=PA212

—

For those people who might not be aware, usage of the DBMS_WORKLOAD_REPOSITORY package requires an Enterprise Edition license and a Diagnostic Pack license, while Statspack requires neither.

Apologies for being off topic here.

I looked at the topic section but I cannot recall seeing any topic on effective measurements of writes and redo logs generated by a query. Tracing is great with tkprof for assorted reads as each process does its own reads. However, the writes and redos will be performed by DBWR and LGWR. I would like to see the volume of physical writes and wait time associated with them.

One way would be to use AWR and take snapshots

EXEC DBMS_WORKLOAD_REPOSITORY.CREATE_SNAPSHOT();
INSERT INTO testwrites
SELECT * FROM tdash;
commit;
EXEC DBMS_WORKLOAD_REPOSITORY.CREATE_SNAPSHOT();

The above is a simple example of inserting 1.5 million rows from one table to another (tdash based on all_objects table + 2 varchar2(4000) columns). Then I can examine the redo size and physical writes. I realise that redo size can also be obtained from autotrace. Are there other alternatives to see the volume of writes generated?

Thanks,

Mich

I buy every book that I review (although in one case the selling site could not ship the book for many months, and then the selling site cancelled my second order for the book), so when first opening the book I expect it to be a book that matches the accuracy and throughness of the better technical books on the market. If I do not see that level of accuracy and throughness, I mention that fact in detail in the review as a cautionary note to anyone else who may have the fortune of reading the same book. On a positive note, with the help of my reviews and the reader comments found in the public comment blog articles that are associated with the reviewed books, some books become much more valuable to the reader. So, thank you to the readers of this blog for helping technical Oracle books fly off the shelves (and hopefully not into the round filing cabinet).

I don’t have the book but do they mentioned multiple tablespace groups?

No mention that I have seen in the book. (That feature is mentioned in the documentation that is referenced above.)

Regarding the use of V$SORT_USAGE, the Oracle documentation indicates that performance view is deprecated as of Oracle Database 9.2. Oddly, even though the documentation indicates that the V$SORT_USAGE view is deprecated, in 11.2.0.2 the V$TEMPSEG_USAGE synonym as well as the V$SORT_USAGE synonym points to the V_$SORT_USAGE view (which of course selects from the V$SORT_USAGE view which selects from the GV$SORT_USAGE view which selects from X$KTSSO).

I think that there are more interesting items in the book’s quote. I will wait for at least 24 hours before sharing my review comments.

Secondly, we can obtain much better performance by avoiding using temp space altogether – one importance of accurate execution plan statistics – or if we have to use temp by trying to avoid multi-pass operations, a topic which might need to branch into workarea policy/sizing.

I don’t have the book but do they mentioned multiple tablespace groups?