ࡱ> :(   D = > ? @ A B C Dp/ 0DTimes New Roman0̱̱~0DArialNew Roman0̱̱~0"@  @@``  @n?" dd@  @@``  &  "! &% '$     *+,S ~1?`f̙999___@8>g4MdMdر~0б ppp@ g4BdBdر~0б ^pgp  <4!d!dL? O I Aphorisms    VDavid Kurtz Go-Faster Consultancy Ltd. david.kurtz@go-faster.co.uk www.go-faster.co.uk8W(-BH!! `0  KPerformance is exactly what the user perceives it to be. No more, no less.LL L K  Performance is  poor when the user s perception does not match their expectation.SS S H  You don t need fancy monitoring software to tell you that a system is performing poorly. [[Z  k:  -Performance Tuning is a search for lost time... . i8  Concentrate on the big picture.    l;  kIt is a capital mistake to theorise before one has data. The Memoirs of Sherlock Holmes, Arthur Conan-Doyle69392 l d3  When you have eliminated the impossible, whatever remains, however improbable, must be the truth. A Study in Scarlet, Arthur Conan-DoyleHb';&  f5  Singularity is invariably a clue. The more featureless and commonplace a crime is, the more difficult it is to bring it home. The adventures of Sherlock Homes, Arthur Conan-Doyle654  |A  It has long been an axiom of mine that the little things are of infinitely important. The return of Sherlock Holmes, Arthur Conan-Doyle6V2V1  m<  To the curious instance of the dog in the night-time. The dog did nothing in the night-time. That was the curious instance. The return of Sherlock Holmes, Arthur Conan-Doyle6|2|1  g6  dIf you don t like the answer, change the question.33 3 x@  Detection is, or ought to be, an exact science. It should be be treated in the same cold and unemotional manner. The Sign of Four, Arthur Conan-Doyle4q%q$  h7  Keep it simple  G  |The problems change from release to release, but the methods by which they are investigated remain the same. Jonathan Lewis Hmm } y>  HYou know my methods, apply them. A Study in Scarlet, Arthur Conan-Doyle4!(!' I J Aphorisms    VDavid Kurtz Go-Faster Consultancy Ltd. david.kurtz@go-faster.co.uk www.go-faster.co.uk8W(-B6!$ /a)n*o+p,q-s/t0u1v2{4~56;<=>?P  ` ___f3f` ___3f` 999MMM>?" dd@,|?" dd@   " @ `"  n?" dd@   @@``@n?" dd@  @@``PR    @ ` ` p>> a Y  4 (  4pF  q0a 4 q 0a2 4 C BUCENG HZI8Qf3? جU,جU,8`T,8`T q0a2 4 C BUCENG HZIDQD? U8U8D`T8D`T 2 4 C BUCENG HZIEQ[? U9U9E`T9E`T dn" 4  G/*d6?@33@@@sf 4 4 Z!gֳgֳ ?  T Click to edit Master title style! !: 4 T$"gֳgֳ ?0  RClick to edit Master text styles Second level Third level Fourth level Fifth level!     S  4  `"gֳgֳ ?`  ?*  4  `"gֳgֳ ?   A*  4  `D#gֳgֳ ?   A*H 4 0޽h? ??0444 ___f3f Fireball.potj     0 8r (  8pF  0  8  0 2 8 C BUCENG HZI8Qf3? جU,جU,8`T,8`T 0 2 8 C BUCENG HZIDQD? U8U8D`T8D`T 9Y 2 8 C BUCENG HZIEQ[? U9U9E`T9E`T " 8  G/*d6?@33@@@sf  8 ZD&gֳgֳ ?`  T Click to edit Master title style! ! 8 Tgֳgֳ ?0 `    W#Click to edit Master subtitle style$ $  8  `Tgֳgֳ ?`  ?*  8  `gֳgֳ ?   A*  8  `gֳgֳ ?   A*H 8 0޽h? ??0888 ___f3f 0 vnP(    T4jJjJ ?S4   _*     TjJjJ ? 4  a*   p  01 ?ss  :  Tgֳgֳ ?q @I  RClick to edit Master text styles Second level Third level Fourth level Fifth level!     S  ZTjJjJ ?S   _*     ZjJjJ ?   a*   H  0zh? ? ̙33 @D( Ć    TjJjJ ?S4   ]*     TjJjJ ? 4  _*     ZtjJjJ ?S   ]*     ZԇjJjJ ?   _*   H  0zh? ? ̙33H< lv( l&6 l l 3 rgֳgֳ ?8`   l # lgֳgֳ ?8  d2 l c ěgֳgֳvd @?UQl$ 0=00 < 2  d2 l c gֳgֳvd @?@UQ< l$ 0>00 < 2  d2 l c gֳgֳvd @? UQ l$ 0?00 < 2  d2 l c Tgֳgֳvd @?@ TP< l$ 0@00 < 2  B l s *޽h ? f3f3f ) t$(  tr t S d0:4   r t S 0:40  H t 0޽h ? f̙33fff > |0(  |x | c $4   x | c $T40  H | 0޽h ? f̙33fff} ;  d(  d d c $4<$ 0  :  d c $d4@<$ 0 :  d Tgֳgֳ ?P ,$D 0 ]That is what users are for.  H d 0޽h ? f̙33fff - $(  r  S D4   r  S 40  H  0޽h ? f̙33fff ,  $( )pM@ r  S 4   r  S $40  H  0޽h ? f̙33fff + 00(  x  c $4   x  c $d40  H  0޽h ? f̙33fff * @$(  r  S d4   r  S 40  H  0޽h ? f̙33fff 2 P$(  r  S 4   r  S T40  H  0޽h ? f̙33fff 6 `0(  x  c $Ի4   x  c $440  H  0޽h ? f̙33fff 1 p$(  r  S T4   r  S 40  H  0޽h ? f̙33fff 0 $(  r  S Ծ4   r  S 440  H  0޽h ? f̙33fff 4 0(  x  c $4   x  c $T40  H  0޽h ? f̙33fff / $(  r  S 4   r  S t40  H  0޽h ? f̙33fff ? `(  `l ` C 4   l ` C 4®40  H ` 0޽h ? ___f3f 5 $( )pM@ r  S ®4   r  S ®40  H  0޽h ? f̙33fffH= tv(  t t 3 rgֳgֳ ?8`   t # lgֳgֳ ?8  d2 t c dgֳgֳvd @?UQl$ 0A00 < 2  d2 t c gֳgֳvd @?@UQ< l$ 0B00 < 2  d2 t c 䙛gֳgֳvd @? UQ l$ 0C00 < 2  d2 t c Dgֳgֳvd @?@ TP< l$ 0D00 < 2  B t s *޽h ? f3f3f0 0 x(  X  C ss   x  S 1:q @I   The availability and response time that a system produces are sum of all the components in the chain. Database - Application Server - Network - PC It takes time for requests to be transmitted over a network, through application server, be processed by the database and returned. There is a certain amount of processing on the client and that takes time too. If any component fails then the system fails. If you go to a cash point and you cant draw money, you don t care whether the central mainframe is down, or the network, or the cashpoint itself is down, or even if it have run out of money. The bottom line is that the system is not available, and you don t get your money. The users can only observe one thing. Their own PC     H  0zh ? ̙333 0 f^ ( )pM@ X  C ss   ^  S q @I   Because when you do the analysis properly you may get an unexpected answer. The answer might be counter-intuitive, it may be so peculiar that you might have rejected it out of hand.  H  0zh ? ̙33; 0 bZ( ػػ0 X  C ss   Z  S $q @I   I rather like Sherlock Holmes, and I have series quotes. It is very easy to jump to a conclusion. But take time to do the analysis properly, otherwise sometimes you get caught.  H  0zh ? ̙33l 8 0 , $ (  X  C ss   $  S 4q @I   If we go back to the cash point analogy the same applies to performance when it is working. What is the total response time. How long to transmit the request to the mainframe, how long for the mainframe to process the request, and how long for the result to be returned. If it takes a minute to do a task, and one component takes 50seconds and one takes 10seconds. Then you stand a better chance of improving the performance by attacking the 50second component. If you went for the 10second component and got it to run in no time at all the process would still take 50seconds. But, it might well be easier to shave 20 seconds off the 50 second component, and so the response time is only 40 seconds. I hate the following phrase, but some people call it  going for the low hanging fruit . I call it common sense. So with a batch process you need to find the longest running SQL statements and work on them. That way you will maximise the effect for the effort.   H  0zh ? ̙33: 0 ~v(  X  C ss   v  S q @I   Usually, concentrate on time not simply i/o or CPU utilisation. It normally all goes together, but sometimes the other measures can be misleading, for instance a high i/o statement can execute very quickly  H  0zh ? ̙337 0 ( S X  C ss     S q @I   By and large, the more simple a thing is, the easier it is to understand, change, test, maintain, upgrade. The more simple a SQL statement the less likelihood there is that the optimiser will take an inappropriate path.  H  0zh ? ̙336 0 `( P'0 X  C ss     S ԁq @I   A change in performance is due to a change somewhere in the system infrastructure (disk subsystem) data volume (cbo statistics) index sql code application Sometimes an approach does not work for whatever reason. Sometimes a programming technique does not scale, or because data volumes, or relative data volumes are different. There are no rules here. No golden nuggets of advice. You are going to have to change something.6r    % H  0zh ? ̙33-< 0 P}( P'0 X  C ss     S tq @I   =Sometimes it is more important that something did not happen.>> > H  0zh ? ̙33 5 0 0\( P'0 X  C ss     S q @I   ^Some of the SQL I see is nothing short of criminal I do not believe in coincidence. If you find something odd, don t discount it as a fluke.   H  0zh ? ̙33@ 0 SKp( @N@ X  C ss   K  S 4q @I   The methodology of tuning is that of the laboratory, that of experimental science. If you want to test or prototype something make you test realistic. You have to make assumptions, but make realistic assumptions. 1. Observe Problem 2. Measure 3. Analyse & Theorise 4. Predict effect 5. Make change 6. Rerun test 7. Measure effect 8. Confirm (not Prove) or Disprove theory if not satisfied go back to the beginning  H  0zh ? ̙33> 0 b(  X  C ss     S q @I   d"Last slide - More Sherlock Holmes.## # H  0zh ? ̙33A 0 w@(  ^  S ss   q  c $q @I   You can find yourself wading through acres of trace files. And often you are looking for a very small fragment of information. So make sure you are monitoring the right attribute.   H  0zh ? ̙33H 0 ZRh(  h^ h S ss   L h c $tĮq @I   A director of a well known company that manufactures monitoring software once said that You don t need fancy monitoring software to tell you that the system is performing poorly. That s what users are for.   H h 0zh ? ̙33I 0 UMp(  p^ p S U   G p # lWW :E   aThis is a rather tongue in cheek presentation that runs through some of my favourite catchphrasesT  H p 0zh ? ̙33KJ 0  x(  x^ x S U    x # lWW :E   I#Mobile phone number +44-7771-760660 $ H x 0zh ? ̙33dK 0 $(  ^  S ss     c $q @I   ROne of the problems with performance is that it can all be a bit touchy feely. People usually say that  the system is slow . So you ask how long they would like it to take, and of course they say no time at all. You need some sort of metric. Performance Level Agreements as a part of the SLA.** H   H  0zh ? ̙33HG 0 (  T R  3 ss     C q @I   The march of so-called progress is inevitable. Software versions change, behaviour changes, bugs appear and disappear. This last bit is a bit like whether the universe will go on expanding for ever or whether they will be enough matter to create enough gravity to cause it to stop expanding and every come together in a big crunch. Perhaps bugs are the dark matter of our universe!  H  0zh ? ̙33r06s=` Jrd`Xf@dZ|`dLTkpHRPVx^z>}\s@"˒x bth{ h\~ ɝ BC,Nf_Bxj2ϦL"JްOh+'0W  ( 4 @ LX`%European Conference 99 - Tree Tuningros Tree Tuning David Kurtz$Tree, Perfromance, Database, OraclerLooking for a way to improve tree performance? Come to this session where youll hear how you can improve performance through a variety of methods, including changes to the tree reading views, replicated security tables, and the database optimiser. This will help you make typical tree reading query execute faster. Bring your tree questions - weve got the answers.eNC:\Program Files\Microsoft Office\Templates\Presentation Designs\FIREBALL.POTl David Kurtz75iMicrosoft PowerPointoso@`EU~@_1"@p/i =@ZEGrGbToM  )* P& &&#TNPP0D & TNPP &&TNPP    --- !---&0&W&--&&- $WjjW- $j}}j - $}}"- $*- $3- $;- $C!- $K%- $R)- $X,- $**\.- $*==*`0- $=PP=b1- $PddPd2- $dwwde2- $ww&&&- &P$&[[\{ 0CUd$p,z3;DLU]zepmdtU{C0{\ZW\L&&-&& &&-&&P$&[[\{ 0CUd$p,z3;DLU]zepmdtU{C0{\ZW\L&&- $WjjW- $j}}j - $}}"- $*- $3- $;- $C!- $K%- $R)- $X,- $**\.- $*==*`0- $=PP=b1- $PddPd2- $dwwde2- $ww&- --&&&O &--&& - $O b bO - $b u ub"- $u  u.- $  9- $  D"- $  P'- $  [-- $  e2- $  o7- $  v:- $ " "|=- $" 5 5"@- $5 H H5A- $H \ \HC- $\ o o\C- $o  o&&&- &N$%T T s   (<!M'\,i2s8z?~EL~SzZs`if\lMq<w({sTROTL&&-&& &&-&&N$%T T s   (<!M'\,i2s8z?~EL~SzZs`if\lMq<w({sTROTL&& - $O b bO - $b u ub"- $u  u.- $  9- $  D"- $  P'- $  [-- $  e2- $  o7- $  v:- $ " "|=- $" 5 5"@- $5 H H5A- $H \ \HC- $\ o o\C- $o  o&- --&&&K!~z&--&&- $K!^!^zKz- $^!q!qz^z.- $q!!zqz=- $!!zzM&- $!!zz[-- $!!zzk5- $!!zzy<- $!!zzC- $!!zzI- $! ! zzN- $ !!z zR- $!1!1zzU- $1!D!Dz1zW- $D!X!XzDzY- $X!k!kzXzZ- $k!~!~zkz&&&- &N$%O!O!n!"#$&(+$.71I4X7d;n?vCzH|LzPvUnYd]XaIe7h$knpsuvwnxOxMxKxOL&&-&& &&-&&N$%O!O!n!"#$&(+$.71I4X7d;n?vCzH|LzPvUnYd]XaIe7h$knpsuvwnxOxMxKxOL&&- $K!^!^zKz- $^!q!qz^z.- $q!!zqz=- $!!zzM&- $!!zz[-- $!!zzk5- $!!zzy<- $!!zzC- $!!zzI- $! ! zzN- $ !!z zR- $!1!1zzU- $1!D!Dz1zW- $D!X!XzDzY- $X!k!kzXzZ- $k!~!~zkz&- --&&&0CqW&--&&- $0C:C:W0W- $:CECEW:W - $ECOCOWEW- $OCZCZWOW- $ZCdCdWZW- $dCoCoWdW- $oCyCyWoW- $yCCWyW"- $CCWW&- $CCWW*- $CCWW.- $CCWW2- $CCWW6- $CCWW:- $CCWW>- $CCWW@- $CCWW@- $CCWW @- $CCWW@- $CCWW@- $C%C%WW@- $%C4C4W%W@- $4CDCDW4W@- $DCTCTWDW"@- $TCcCcWTW&@- $cCsCsWcW)@- $sCCWsW-@- $CCWW2@- $CCWW6@- $CCWW9@- $CCWW=@- $CCWWB@- $CCWWH@- $CCWWM@- $CCWWR@- $CCWWW@- $CCWW\@- $C!C!WWa@- $!C.C.W!We@- $.C;C;W.Wj@- $;CHCHW;Wo@- $HCVCVWHWu@- $VCcCcWVWz@- $cCpCpWcW~@- $pC}C}WpW@- $}CCW}W@- $CCWW@- $CCWW;- $CCWW5- $CCWW1- $CCWW -- $CCWW')- $CCWW/%- $CCWW6!- $CCWW=- $CCWWD- $CCWWK- $CCWWS- $CCWWZ - $CCWWa - $CCWWh- $CCWWo- $CCWWv- $C&C&WW~- $&C.C.W&W- $.C6C6W.W- $6C;C;W6W- $;C@C@W;W- $@CECEW@W- $ECJCJWEW- $JCOCOWJW- $OCUCUWOW- $UCZCZWUW- $ZC_C_WZW- $_CdCdW_W- $dCiCiWdW- $iCoCoWiW- $oCqCqWoW&&&- &,$9C6D3F1I0L0L1P3R6T9UgUjTmRnPoLoLnImFjDgC&&-&& &&-&&,$9C6D3F1I0L0L1P3R6T9UgUjTmRnPoLoLnImFjDgC&&- $0C:C:W0W- $:CECEW:W - $ECOCOWEW- $OCZCZWOW- $ZCdCdWZW- $dCoCoWdW- $oCyCyWoW- $yCCWyW"- $CCWW&- $CCWW*- $CCWW.- $CCWW2- $CCWW6- $CCWW:- $CCWW>- $CCWW@- $CCWW@- $CCWW @- $CCWW@- $CCWW@- $C%C%WW@- $%C4C4W%W@- $4CDCDW4W@- $DCTCTWDW"@- $TCcCcWTW&@- $cCsCsWcW)@- $sCCWsW-@- $CCWW2@- $CCWW6@- $CCWW9@- $CCWW=@- $CCWWB@- $CCWWH@- $CCWWM@- $CCWWR@- $CCWWW@- $CCWW\@- $C!C!WWa@- $!C.C.W!We@- $.C;C;W.Wj@- $;CHCHW;Wo@- $HCVCVWHWu@- $VCcCcWVWz@- $cCpCpWcW~@- $pC}C}WpW@- $}CCW}W@- $CCWW@- $CCWW;- $CCWW5- $CCWW1- $CCWW -- $CCWW')- $CCWW/%- $CCWW6!- $CCWW=- $CCWWD- $CCWWK- $CCWWS- $CCWWZ - $CCWWa - $CCWWh- $CCWWo- $CCWWv- $C&C&WW~- $&C.C.W&W- $.C6C6W.W- $6C;C;W6W- $;C@C@W;W- $@CECEW@W- $ECJCJWEW- $JCOCOWJW- $OCUCUWOW- $UCZCZWUW- $ZC_C_WZW- $_CdCdW_W- $dCiCiWdW- $iCoCoWiW- $oCqCqWoW&- --&&&--yH-- wwgw - Times New Romanwgw - f.2 c Aphorisms$*.--qls-- fTimes New Romanwgw - .2 A| David Kurtz   . .'2 |Go-Faster Consultancy    . . 2  Ltd  . . 2 2. . .2 |david.  . .2 kurtz . .2 6@go-faster.co.uk#    . .$2 |www.go-faster.co.ukc     .&9#eO&--vG-Nd#9yI-Lb#9L-Ka$:Q-I_&<W-H^'=^-G](>f!-E[*@n$-DZ+Aw'-CY-C~*-BX.Dޅ,-AW/E.-?U1G/->T2H1-=S3I2-;Q5K2-:P6L---&& &9`e&--vG-d`9yI-b`9L-aa:Q-_c<W-^d=^-]e>f!-[g@n$-ZhAw'-YjC~*-XkDޅ,-~WlE.-|UnG/-{ToH1-zSpI2-xQrK2-wPsL---&& &9e&--vG-d9yI-b9L-a:Q-_<W-^=^-]>f!-[@n$-ZAw'-YC~*-XDޅ,-WE.-UG/-TH1-SI2-QK2-PL---&& &9e &--vG- d9yI- b9L-a:Q-_<W-^=^-]>f!-[@n$-ZAw'-YC~*-XDޅ,-WE.-UG/-TH1-SI2-QK2-PL---&& --"Systemwf  -&TNPP &՜.+,D՜.+,`   $ ,4<D L (PeopleTools and Technology/DB PlatformsOn-screen Show  PeopleSofthұi1 Times New RomanArial Fireball.pot AphorismsNo Slide TitleNo Slide TitleNo Slide TitleNo Slide TitleNo Slide TitleNo Slide TitleNo Slide TitleNo Slide TitleNo Slide TitleNo Slide TitleNo Slide TitleNo Slide TitleNo Slide TitleNo Slide TitleNo Slide Title Aphorisms  Fonts UsedDesign Template Slide Titles 0s{ _PID_GUID _PID_HLINKSTelephone NumberAN{7E4EF55D-3F28-11D4-9F4C-00105AF9AD23}A,0257,2,Custom Topic A258,3,Custom Topic B259,4,Vital Statistics260,5,Favorite Links to Visit257,2,Custom Topic A258,3,Custom Topic B259,4,Vital Statistics260,5,Favorite Links to Visit+44-118-9521320Root EntrydO)0 @Current User+SummaryInformation(YWPowerPoint Document(ұ#_㮱 David Kurtz  !"#$%&'()*+,-./0123456789:;<=>?@ABCDEFGHIJKLMNOPQRSTUVWXZ[\]^_`abcdefghijklmnopqrstuvwxyz{|}~Root EntrydO)Current UserSummaryInformation(YWPowerPoint Document(ұDocumentSummaryInformation8