This discussion is archived
1 2 Previous Next 23 Replies Latest reply: Oct 29, 2013 12:50 AM by Girish Sharma Go to original post RSS
  • 15. Re: Key Performance Related Parameters in Oracle 10g
    sybrand_b Guru
    Currently Being Moderated

    With due respect, this is not correct.

    Did ADDM ever *correct* anything automatically at all?

    ADDM is a very *coarse* mechanism, which *doesn't know your DATA* and whose recommendations are often to

    - defragment segments

    - increase memory

     

    It is an aide, nothing more and nothing less.

    And you are blinded by Oracle marketing speech.

    You believe all the exaggerations they publish. Sad!

    --------------

    Sybrand Bakker

    Senior Oracle DBA

  • 16. Re: Key Performance Related Parameters in Oracle 10g
    sybrand_b Guru
    Currently Being Moderated

    The link you posted is from a document about Oracle Fusion Middleware.

    What has Fusion Middleware tuning to do with tuning Oracle RDBMS?

     

    -----------------

    Sybrand Bakker

    Senior Oracle DBA

  • 17. Re: Key Performance Related Parameters in Oracle 10g
    Nikolay Savvinov Guru
    Currently Being Moderated

    Hi,

     

    GirishSharma wrote:

     

    >Oh no. Really? How suprising. That's probably the first case in the history of mankind when a manufacturer claims his product can do things that it can't.

     

    It depends, who is manufacturer, how his products are recognized / rating in the industry etc. 

     

    No, it doesn't. Nissan is a great manufacturer, it has world-wide recognition and very high ratings in the industry. Nevertheless their cars can't do even 10% of what their commercials promise, like magical computerized dogs appearing out of nowhere and pulling the car from an abyss.

     

    When official docs are claiming that ADDM can locate the root causes of performance problem, it means YES it can (if it is properly used, configured and understood) and if it is not, then anyone can file a doc bug with providing a complete test case.

     

    Even though, I am not providing/knows a test case that how ADDM will locate performance problems but since it is written in the docs and have not seen a test case which shows that ADDM is not able to locate performance issue by providing test case, then I have to believe upon it.

     

    Great scientist (Oracle Engineers) have invented many surprising things/theory/algorithm for mankind though and mankind always have being surprised by their inventions many time.

     

    Regards

    Girish Sharma

     

    Once I went to see a presentation by Jonathan Lewis. After the presentation I asked him to sign a book ("Cost Based Oracle Fundamentals"). He did, and he also wrote these wonderful words: "just because it's printed, doesn't mean it's true".

     

    I highly recommend that you take on his advise as well.

     

    Best regards,

    Nikolay

  • 18. Re: Key Performance Related Parameters in Oracle 10g
    Girish Sharma Guru
    Currently Being Moderated

    >Did ADDM ever *correct* anything automatically at all?

    Neither I nor docs are saying like that.  ADDM locates root cause of performance problems.  If you or anyone have objection on this sentence (ADDM locates root cause of performance problem) then how and why you are saying like that ?  Can you or anyone provide any test case to calumniate it ?  Even though as I said earlier, I am also not able to find how ADDM locates root causes of performance problem, but since it is written in docs so I shall believe on it until and unless I don't sees it prove false.

    >The link you posted is from a document about Oracle Fusion Middleware.

    I know it, but still it saying about those performance related parameters which OP is looking for and I am 100% agree with those parameters which have been mentioned in the above doc link.

     

    @Nikolay

    So you mean Nissan is having same credibility and popularity in the Auto Industries as Oracle have in IT industry ?  Nissan and Oracle both are same ?  Are you sure you are comparing "eligible candidates" ?  I am neither Nissan's nor Oracle's Sales representative, but I doubt you are comparing eligible candidates.

     

    > "just because it's printed, doesn't mean it's true".

    Again, it depend who have written it.  How many time writer have prove right/wrong.  Writer is human, he can commit mistake but it does not mean he/she is not eligible for trust just because of 7-10% mistakes.  If I gets 90% correct information from him then I can correct myself for rest 10% and can communicate to writer to make necessary corrections (if any).

     

    But, still I am surprised that you both are saying that ADDM do not locates performance problem, it just provide "template" kind of recommendations, it is waste of time  to search performance problem by ADDM, then Oracle don't knows about it ?  Why should I believe upon your saying ?

     

    Regards

    Girish Sharma

  • 19. Re: Key Performance Related Parameters in Oracle 10g
    jgarry Guru
    Currently Being Moderated

    John hit on the major ADDM problem for this thread:  OP is on Standard Edition.  I know I'm not alone in thinking it is a Bad Thing (tm) that Oracle doesn't even let you license performance tools.

     

    As far as the others, one can quibble about details, but they are essentially correct:  you still have to apply human brainpower to the tools.  Especially when they go wrong because there are more variables than they know about and/or the business requirements have different goals.

     

    I do somewhat disagree with the idea that nothing should be looked at; even in the modern times of 10gR2, the defaults for redo and memory may be way off, and should at least have a sanity check.

     

    Regarding Nissan:  Wife loved her Murano, until it proved itself unreliable.  In particular, the transmission showed signs of being extremely unhappy, so we had to dump it cheap.  Nearly a year later, they sent me a letter saying there's nothing wrong with the transmissions, but they're extending the warranty anyways.  No more Nissans here.  

  • 20. Re: Key Performance Related Parameters in Oracle 10g
    Nikolay Savvinov Guru
    Currently Being Moderated

    Hi,

     

    > "just because it's printed, doesn't mean it's true".

    Again, it depend who have written it.  How many time writer have prove right/wrong.  Writer is human, he can commit mistake but it does not mean he/she is not eligible for trust just because of 7-10% mistakes.  If I gets 90% correct information from him then I can correct myself for rest 10%

     

    Well then let me put it this way: Oracle documentation is (at least) 90% correct, but unfortunately the claims they make about ADDM being able to find root causes of performance issues doesn't fall into this category. I

     

     

    But, still I am surprised that you both are saying that ADDM do not locates performance problem, it just provide "template" kind of recommendations, it is waste of time  to search performance problem by ADDM, then Oracle don't knows about it ?  Why should I believe upon your saying ?

     

     

     

    I never said that ADDM is always a waste of time. Often, it can provide information that is helpful for identifying the root cause of an issue. However, the irony is that if you have enough knowledge of performance tuning to interpret ADDM data correctly, then you don't really need ADDM. You have AWR, ASH, SQL trace, dbms_xplan and many other tools which get you to the point faster.

     

    The main problem with ADDM is that in wrong hands (and it always seems to be in wrong hands) it's an endless source of "make work". It doesn't seem to have the option of saying "your database is fine, sit back, relax and enjoy it while it lasts". It contributes to the spread of the "compulsive tuning disorder" in the database community more than any other tool.

     

    Another thing I don't like about ADDM is that it makes so easy for incompetent DBAs to hide behind its recommendation. A typical scenario: a performance issue happens, the DBA on shift doesn't have a clue, so he runs ADDM, which basically puts the finger to the SQL with highest elapsed time around the time of incident, and calls it the root cause. You spend days tuning this query only to find out that nobody cares about its performance, it has always been as slow as this, and that the root cause of the issue was something else. If there was no ADDM, the incompetent DBA would have to admit that he doesn't have any idea what was going on, and you would have had a decent shot at performing root cause analysis yourself.

     

     

    Best regards,

       Nikolay

  • 21. Re: Key Performance Related Parameters in Oracle 10g
    Girish Sharma Guru
    Currently Being Moderated

    >you have enough knowledge of performance tuning to interpret ADDM data correctly, then you don't really need ADDM.

    Wrong because when I am feeling pain in my legs, doctor suggested me to go for x-ray, I am not able to understand x-ray, but it helps doctors very much.  Doctor knows the reason of cause of pain with x-ray.  X-ray (ADDM) is playing its role for what it mean.


    Apart from many other reasons which you have provided, suppose I writes to Oracle that ADDM is not "good" tool for performance issues, Oracle will simply ask me "Can you proof it by showing sqlplus cut/paste? What performance issue you are addressing and what recommendations you got by ADDM?  This is what I want to see that in this scenario ADDM is not providing correct recommendation(s).


    Can you help me to simulate couple of performance problem and see how ADDM recommends on my issue which I have created deliberately please?


    I am not saying that only ADDM is the ultimate and super-best tool, there are many other which works very well in different scenario and environments, but when I see around 1000 times recommendations by many ACEs, Experts, Gurus to use ADDM for performance issues, and now I am seeing opposite one, then I am getting confused and this confusion is now looks like fight/arguments!


    I think our discussion is now far away from OP's question, I should not continue here, otherwise it looks like hijacking.

     

    Regards

    Girish Sharma

  • 22. Re: Key Performance Related Parameters in Oracle 10g
    Nikolay Savvinov Guru
    Currently Being Moderated

    Hi,

     

    GirishSharma wrote:

     

    ...

    I am not saying that only ADDM is the ultimate and super-best tool, there are many other which works very well in different scenario and environments, but when I see around 1000 times recommendations by many ACEs, Experts, Gurus to use ADDM for performance issues, and now I am seeing opposite one, then I am getting confused and this confusion is now looks like fight/arguments!

     

     

    Look at it this way: if ADDM is indeed so good that it sees right through performance issues like an X-ray, then people like Tanel Poder, Randolf Geist or Jonathan Lewis, who make their living troubleshooting Oracle performance and writing (talking) about it, must use it every day.

     

    Don't you find it strange that they haven't written a single article, a single blog entry, a single presentation about ADDM?

     

    Best regards,

    Nikolay

  • 23. Re: Key Performance Related Parameters in Oracle 10g
    Girish Sharma Guru
    Currently Being Moderated

    I don't know why they developed their own tools/scripts when there is ADDM and many more other provided by software vender.  What performance issues they got and ADDM failed to locate their issues.  Neither I found any blog entry for it nor any explanation in forum that I was doing this and this was ADDM recommendation, I am happy with ADDM and shall try to simulate performance issue and see how ADDM replies me.  Obviously if I gets "template" recommendation, I shall open a new thread (but don't know how much it will take time though!)

     

    Regards

    Girish Sharma

1 2 Previous Next

Legend

  • Correct Answers - 10 points
  • Helpful Answers - 5 points