Re: Oracle Performance Tuning - Tips & Techniques (Possible Error found) ?

You are correct in your fix for this equation. The db block gets + consistent gets = logical reads which include the physical reads. The total number of reads in both equations should be 100,000 in both equations. Thanks for this fix ... I missed this one. As far as fallible Oracle experts go ... goes with the territory of sharing the knowledge ... it was certainly one of the fears that I had in writing the book was that I knew that it wouldn't be perfect given the ground that I had to cover... after 2 years of working on it trying to get it to be perfect (the curse of the DBA mind) they finally cut me off on working on it ... there are a few mistakes (shared later in this missage) and it is the price of sharing knowledge...you just can't get it 100% ... but through the great newsgroups like this I'll find a lot of them and the next revision will be better ... but I'm sure still not perfect. After over a decade of working with Oracle and tuning it is still a challenge to stay educated. One final note on the book ... it's not written like a novel ... more like I think...bursts of tips that hopefully you will find one that makes a huge impact for you ... if you do ... share it here so that others can be pointed to the tips that helped most.

One other note that a reviewer brought up is that "db_writers" from Oracle 7 is "dbwr_io_slaves" in 8 for non-asynch systems (not all) where db_writers in Oracle7 is now "db_writer_processes" in 8 if your O.S. takes advantage of async. io properly...although I got killed by the reviewer for missing this, the parameter didn't exist on the version of Oracle8 at the time that I wrote the book...what can you do...perfection is elusive.

I also plan to setup a website with any corrections (Kevin Loney did this for his / a great idea)... and also 8i additions that they wouldn't allow (they cut me off at 900 pages complaining that I was only supposed to write 300).

Other corrections for your information (I feel pain just having to list these):

Page 16: Should be AM instead of PM for first 5 listed times (ie Get everthing into memory)
Page 51: USE_HASE should be USE_HASH.
Page 74: The second '05-JUL-1998' needs a TO_DATE on it to add the .99999 (giving you 23hrs 59min 59sec)
Page 185: #12 should clarify that DB_FILE_MULTIBLOCK_READ_COUNT is for full table scans.
Page 364: Missing an * on the hint (I actually corrected this one in final reviews)
Page 538: A Clarification: The NOLOGGING is in addition to UNRECOVERABLE in Oracle8... NOLOGGING causes indefinate NOLOGGING of information whereas UNRECOVERABLE is for the operation only (such as a create table or index).

That's all I have right now...I'll try to post any new finds ... also for the people that note some of the repetition ... that was intentional as I have always hated books that said "see page xyz" for something... I always hated having to turn to another page... so I repeated "a few" things so that you don't have to do that and so that each chapter stood by itself for the most part (it was written a lot for people without a lot of time that could read one chapter at a time and the chapter would stand by itself).

Lastly, I think that chapter 9 & 17 were by far the most painful (especially 9) and will be the most helpful... I really had to fight for 9 (for them to keep it in due to length) and it was a killer to write and very insightful if you read it carefully.

Although I can never get back the pain that I put into this one ... the goal was always that every reader would find one tip that would "knock their socks off" when they tuned a query from hours to minutes or seconds and that would make all of the pain worth it. I've got a lot of people that have sent me some great ones ... so the journey was worth it. If you're thinking of writing a book you might look for something buried at the end of Appendix A ...(they took this out twice on me and I slipped it in here :) on why NOT to write a book). Also, buried at the end of Appendix B is a great description of the creation of the v$ and x$ tables that I got in at the last minute (it was the only place they would put it). On the TUSC (www.tusc.com) website is presentation on Tuning that includes some of the stuff that was too late on Oracle8i including examples on materialized views and function-based indexes. If you read this far and email me (niemiecr_at_TUSC.com) I'll send it to you directly.

Thanks for the input and sharing the knowledge...I'll keep working to make it better.

Rich Niemiec
Oracle Performance Tips and Techniques, Oracle Press ISBN: 0078824346 Received on Wed Aug 04 1999 - 11:44:43 CDT