OReilly SQL Tuning Dec 2003 ISBN 0596005733

Foreword
Duringmyyearsasadeveloper,andlaterasadatabase
administrator,theonlyperformancetuningareainwhichI
developedanyexpertiseandhadanylevelofsuccesswasthat
oftuningindividualSQLstatements.Becauseofthat
background,IwasveryinterestedwhenDanTowcametome
withhisideaforabookonSQLtuning.
TheproblemwithSQLtuning,atleastfrommyown
perspective,isthatitwasoftenreasonablyeasytopinpoint
badlyperformingSQLstatementsthatneededtobetuned,and
itwasreasonablyeasytodeterminetheexecutionplan
currentlybeingusedforthosebadlyperformingstatements,but
thencamethehardproblem:findingabetterexecutionplan,to
makeanoffendingSQLstatementexecutefaster.Manyisthe
timeI'vestaredatapoorlyperformingSQLstatement,
reviewingitspoorlyperformingexecutionplan,andwondered
justwhatIshoulddonext.Forthatmatter,wasthereevenany
roomforimprovement?Perhapstheexecutionplanthat
performedsobadlyfromauser'sperspectivewas,infact,the
bestpossibleexecutionplan.PerhapsIwaswastingmytime
tryingtoguessatabetterone.
There,Iusedthatwordguess,andthat'sattheheartofwhat
sometimesmadetuningSQLstatementsafrustratingactivity.It
allcamedowntomylookingataSQLstatementandtryingto
guessatabetterplan.Ofcourse,I'dattempttofactorinmy
experience,myintuition,andmyknowledgeofthedatabeing
queried,andI'dpullouttipsandtricksthatI'dreadaboutin
booksandmagazinearticles,butintheendI'dmakeaguess,
tryanewplan,makeaguess,tryanewplan,andsoforth.I'd
stopwhenoneoftwothingshappened:
Igotluckyandguessedaplanthatwasagood-enough

improvementovertheoldplantosatisfymyclient.
Iranoutofideas.
ItalwaysbotheredmewhenIranoutofideas,becauseInever
knewforsurewhetherthecurrentplanreallywasoptimalor
whetherIwasjusttoothick-headedtointuitabetterperformingexecutionplanthanthecurrentone.Tobehonest,I
wasalwaysprettyhardonmyself,chalkingupanyfailureto
improveaSQLstatementasapersonalinadequacy.
Dandoesn'tguess.Letmeboldfacethatandunderlineit.He
doesn'texecuteaniterativeguess-looplikeIdid,tryingone
ideaafteranotherinthehopeofstumblingacrossan
improvement.Instead,Danusesaninnovativeand
mathematicallybaseddiagrammingmethodtoderivethe
optimal,ornear-optimal,executionplanforaSQLstatement.
Thenheputsthatplanintoeffect.Andthat'sit.There'sno
guesswork,andthere'snouncertaintyastowhetherfurther
improvementispossible.
Atfirst,IwasskepticalofDan'sapproach.ButthemoreIread
hischapters,themoreIbegantoseethelogicbehindhis
method.Mathdoesn'tlie,andexperienceisagoodindicator.
Danhasover10yearsoftuningexperience,andhe'sbeenvery
successfulusingthemethoddescribedinthisbook.
TherearethreelegsthatanySQLtuningeffortneedstostand
on.YouneedtoknowhowtoidentifyabadlyperformingSQL
statement.Youthenneedtobeabletoseetheexecutionplan
usedforthatstatement.Finally,youneedtosomehowcomeup
withanimprovedplan.I'mconvincedthatDan'smethod
representsaviablethirdlegonwhichyourSQLtuningefforts
canrest.Readhisbook,applyhismethod,andsaveyourself

hoursofwonderingwhattodonext.
JonathanGennick,author,editor,OracleDBA