1 2 3 4 5 Previous Next 63 Replies Latest reply: Jan 22, 2008 9:52 AM by 314771 Go to original post RSS
      • 45. Re: DB Block Size
        153119
        Evilghostrider. Most people do not contribute anything useful. You are no exception.
        Most people here also don't come here for help, but because they are too lazy to look up anything, or to read documentation.
        If you actually would follow this forum, you would recognize 99.9 percent of all questions are of the RTFM type. It is THAT what I am against.
        This is a discussion forum.
        It is not
        - a questions and answers forum
        - a 'Help, I am too lazy to do my own work. I need the steps. Urgent' forum
        - a 'Help, I want you to do my work for free'
        Just look at the thread about Capacity Planning which is going on for almost 2 years, and you will know.
        You don't understand what the nature of this forum.
        It has gone to the abyss thanks to a bunch of lazy newbies, and thanks to a self-proclaimed 'guru' who is misusing it to promote his own books.

        --
        Sybrand Bakker
        Senior Oracle DBA
        • 46. Re: DB Block Size
          70140
          To add another view to your comment above;

          It might be noteworthy to let us know the kind of application(s), using this database. I know it may seem insignificant, but WE all know that the suggestions proffered can only make the most applicative sense if the sort of application(s) and its corresponding database use - (S U I D), is understood.

          Hence, we may be able to advice better...and what sort of initial installation was done - OLTP, General et.c

          If all has been considered before, please forgive, but I did not see it mentioned in my half-hearted review of this VERY interesting subject ( Too many comments for me to follow)
          • 47. Re: DB Block Size
            617226
            I would argue that it has gone to the abyss as you call it because of people like you :)

            - a questions and answers forum
            Really, most forums I know of usually involve some kind of questions and answers
            - a 'Help, I am too lazy to do my own work. I need the steps. Urgent' forum
            Or maybe these people are new to Oracle and look to this forum for genuine help because they need it, its a shame there are people like you who make there visit here there last.
            - a 'Help, I want you to do my work for free'
            I don't think anybody expects anyone else to do there work, please accept that sometimes people just want help.
            • 48. Re: DB Block Size
              611433
              Sigh!!.....
              Sigh!!......

              I am sorry if my question caused this dispute, which in the end leads to nothing except bitter thoughts.

              I am still not sure weather I should raise the issue of increasing the block size to my managers or should I talk to my application user to sort their queries.
              • 49. Re: DB Block Size
                Richard Foote
                Hi Don

                Just a few questions, points if I may.

                Why does your quote from the 11g documentation have nothing to do with multi-block sizes ? If anything, it's a discouragement against them as the KEEP and RECYCLE pools must be the default block size, right ?

                Where in the OP's statspack report do you see any justification for multiblock sizes ?

                I consider myself to be a DBA and yet I most certainly don't use multi sized blocks, therefore by definition it's not "a tried and true technique that every working DBA uses".

                It's incorrect to suggest that "All else being equal, bigger blocks = less I/O overhead". In say a typical OLTP environment, how can say reading 1 32K index root block + 1 32K index leaf block + 1 32K data block be less overhead than 1 8K index root block + 1 8K index leaf block + 1 8K table block ?

                I listed a range of reasons why multiblock sizes may not be such a good idea and Greg Rahn showed in a nice example of how larger block sizes made no difference (or matters worse) in this previous thread: Index blocksize

                In the thread I showed you how a larger multiblock size makes no difference to an Index Fast Full Scan, one of your stated advantages of using larger block sizes. Can you explain why a larger block size will improve the performance of an Index Fast Full Scan if Oracle is actually performing the same I/Os regardless ?

                Note Greg also didn't see where the performance benefit was either, so I guess that makes at least two of us ;)

                Cheers

                Richard Foote
                http://richardfoote.wordpress.com/
                • 50. Re: DB Block Size
                  153119
                  Evilghostrider,

                  99.9 percent of all questions here can be LOOKED UP in the manual.
                  If people want help, there is help, ONLINE, IN THE MANUAL!

                  Most people here don't WANT to read the manual.
                  Most people here want EVERYTHING on a SILVER SPOON on THEIR DOORSTEP.

                  It is a SHAME there are so many people who can't be bothered to look up the manual, and it is A SHAME you reproach me for saying so.
                  So far, you have contributed nothing but FLAMES.
                  If you don't want to help people, please GO AWAY.

                  --
                  Sybrand Bakker
                  Senior Oracle DBA
                  • 51. Re: DB Block Size
                    588568
                    Hi Olu: The testing I did was on a scratchpad database ages back, and it's no longer on my home server. As I said, I was just tinkering about, and didn't have time (then) to progress with my testing. All I can say is that multiple blocksizes were "good", "indifferent", or "bad". There was nothing terribly obvious about what the common factor was. Sorry! Hoping Don or A.N.Other might be able to offer us all a bit of decent evidence.
                    • 52. Re: DB Block Size
                      588568
                      Written to your soon-to-be-litigating customers, yet?
                      • 53. Re: DB Block Size
                        153119
                        'MOST DBA's --> Don

                        This is also known as the 'authority argument'

                        --
                        Sybrand Bakker
                        Senior Oracle DBA
                        • 54. Re: DB Block Size
                          611433
                          sybrandb,

                          You are right, but the question is " Which Book"? I have read some books that give completely different opinions than that of the Oracle Documentation.

                          So we, the so called lazy newbies, come to people like you who have more knowledge, more experience and know what effect certain solutions might have.

                          Books dont teach you everything, you have to have experience below your belt to sort your some problems. As I work in a live envoierment, I just cannot afford to take risks.
                          • 55. Re: DB Block Size
                            588568
                            Richard: By switching tables and indexes to different blocksized tablespaces, I have seen benefits in response times... and also terrible performance. I wonder if the reason for multiple blocksizes being (it seems) universally unused for tuning is that it's too much like guess work as to whether any particular table or index will, in fact, benefit from it, without trying and testing from scratch each time?
                            • 56. Re: DB Block Size
                              617226
                              quote:"If you don't want to help people, please GO AWAY."
                              Hold on a minute ago you said this was not a help forum!!!

                              It is a SHAME you take such a negative stance on everything.

                              At first I thoiught you were a dude but you gotta be chick with PMT, im I right?
                              • 57. Re: DB Block Size
                                chris_c
                                This dispute has been going on for a long time and will be around for a long time to come, its nothingto do with the question an unfortunatly these disputes spill over form the technical to the personal sometimes.

                                I would start at the application level, start with the most expensive sql and work from there, if the application is written in house finding poorly performing sql and fixing it is often the best way to improve performance, even if the sql is unmodifiable it helps to understand the application.

                                also I would look at taking snapshots more frequently the one you posted covers 130 minutes, 10-15 is usually a better interval to attempt to tune espesially if you are getting variations in performance (i.e. sometimes good sometimes bad).

                                Also has the application always run against a rac database or recently been moved from a single instance?

                                Chris
                                • 58. Re: DB Block Size
                                  70140
                                  Thanks HarryP.

                                  I hope Tiger wont be contemplating to scream out loud or pull out his hair now....

                                  I find the subject interesting, NOT only because its getting 'self-styled' experts hot under the collars (pardon my expression...but look through the comments), but also because as it concerns Performance Tuning...et.c, a grayness still pervades what is 'Correct' or 'Incorrect'.

                                  It is my opinion and I think that of oracle, that a DB can only perform optimally if the sort of overall application usage is determined before creating that database, making sure to align the initialization parameters to suit.

                                  I was also of the hope that we will be able to arrive at some kind of 'concensus' via comments and contributions, rather than bickering and name-calling (a la Obama/Clinton style)
                                  • 59. Re: DB Block Size
                                    Richard Foote
                                    Hi Harry

                                    I believe the reason is simply there's generally so much lower hanging fruit in an environment from a tuning perspective that the minor benefits you may get from very specific usages of multi sized blocks just isn't worth the extra administration and the associated risks of doing more harm than good.

                                    Cheers

                                    Richard Foote
                                    http://richardfoote.wordpress.com/