REP-1993-120 - Sen s it iv ityA na ly s iso fF requencyC...

Info iconThis preview shows pages 1–3. Sign up to view the full content.

View Full Document Right Arrow Icon

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: Sen s it iv ityA na ly s iso fF requencyC oun t ing TheodoreJohn son D ep t .o fC om pu terandIn form at ionSc ience U n ivers ityo fF lor ida A b stract M anydatabaseop t im izat ionact iv it ies ,suchasp re fetch ing ,datac lu ster ingandpart it ion ing ,and buera llocat ion ,dependonthedetect iono fhotspotsinaccesspattern s .W h ileadatabasedes ignercan insom ecasesu sespec ia lknow ledgeabou tthedataandtheu serstop red icthotspots ,ingenera lone m u stu sein form at ionabou tpastact iv itytop red ictfu tu react iv ity .H ow ever ,a lgor ithm sthatm akeu se o fhotspotspay litt leatten t iontothew ay inw h ichhotspotin form at ion isgathered ,ortothequa lityo f th isin form at ion .Inth ispaper ,w ep resen tam ode lforana lyz inghotspotest im atesbasedon frequency coun t ing .W ep resen tanum er ica lm ethod forest im at ingthequa lityo fthedata ,andaru le-o f-thum b . W endthatif b o fthere ferencesarem adetothehottest a o fthe N data item s ,thenoneshou ldp rocess N a d ( a ) = ( ba ) e re ferences . In troduct ion M anydatabaseop t im izat ionact iv it iesrequ iretheiden t icat ionandc lass icat iono f ho tspo ts ,orreg ion so f thedatabasew ithanexcep t iona llyh ighaccessfrequency .Insom ecases ,adatabasedes ignercan iden t ify hotspotsbasedonspec ia lknow ledgeo fthedataandtheu sers .H ow ever ,oneisu sua lly forcedtop red ict fu tu rere ferenc ingpattern sbasedonpastbehav ior .A comm onapp roach istou se frequencycoun ting .The num bero fre ferencestotheob jectsinthesy stem arecoun tedoveraper iodo ft im e .O b jectsw ithah igh frequencycoun tarelabe led/hot" ,andtheotherob jectsarelabe led/co ld" . F requencycoun t ingw illaccu rate ly iden t ifythehotob jectsinthedatabaseifthere ferencepattern is stat ionaryand ifenoughre ferencesareco llected .H ow ever ,oneu sua llyw an tstom akedatabaseorgan izat ion orre-organ izat iondec is ion sbasedon lim iteddata .Theear lierthehotspotsaredetected ,thesoonerthe databasecanbeop t im izedandthecorrespond ingperform ancebenetsob ta ined .Inadd it ion ,re ference pattern sareu sua llynotstat ionary ,anddec is ion sbasedonsta ledatacandegradeperform ance .S inceoneis forcedtom akedec is ion sbasedon lim iteddata ,am ethodo feva luat ingthequa lityo fthehotspotin form at ion isneeded . Inth ispaper ,w eexam inetheperform anceo ffrequencycoun t ing .Theinpu ttothefrequencycoun t ing a lgor ithm isa re ferencestr ing com posedo fthesequenceo fob jectre ferencesthatthesy stem generates . W eacknow ledgethesupporto fUSRAgran t#-.P arto fthew orkw asperform edwh ileTheodoreJohn sonw asan ASEESumm erF acu ltyF e llowattheNASA 'sN at iona lSpaceSc ienceD ataC en ter W eassum etheIndependen tR e ferenceM ode l(IRM ) ,sothateachre ferenceisindependen tand iden t ica lly d istr ibu ted .G iventhelengtho fthere ferencestr ingandthed istr ibu t iono fre ferences ,w eca lcu latethe qua lityo fthehotspotest im ate .O nem easu reo ftheest im atequa lity isthep robab ilitythatahotob ject islabe led/hot" .A secondm easu...
View Full Document

Page1 / 19

REP-1993-120 - Sen s it iv ityA na ly s iso fF requencyC...

This preview shows document pages 1 - 3. Sign up to view the full document.

View Full Document Right Arrow Icon
Ask a homework question - tutors are online