setupnokia

OntheEnergyConsumptionofDesignPatternsChristianBunseUniversityofAppliedSciencesStralsundZurSchwedenschanze1518435StralsundSebastianStiemerUniversityofAppliedSciencesStralsundZurSchwedenschanze1518435StralsundMarch29,2013Introduction.
Energyisoneofthemostlimitingfactorsforinformation&communicationtechnologiesingeneraland,morespecicallyformobiledevicessuchasSmartphones.
Inmostapplicationscenarios,mobiledevicesdonothaveapermanentpowersupplybutuserechargeablebatteries.
Duetotheincreasinghard-wareperformanceandotherdevicepropertiesenergyrequirementsincreasefurther.
However,softwareuti-lizeshardwareandthereforedirectlyaectstheenergyrequirementsoftheentiresystem.
Energy-awaresoftwaredevelopment,energy-awarealgorithmsandenergy-awaresensorsubstitutionareonlythreeexamplesforrecentresearchthattrytoreduceenergyrequirementsbyoptimizingthesoftwareratherthanthehardware.
Energyconsumptionisanimportantsystemproperty,thathasalreadytobeaddressedintheearlystagesofdevelopment.
Inturn,thisrequiresknowledgeonbest-practicesandstructuresfordevelopingenergy-ecientsoftwaresystems.
Following[1],patternsplaymanyrolesinsoftwaredevelopment:theyprovideacommonvocabulary,re-ducesystemcomplexity,constituteabaseforbuildingreusablesoftware,andactasbuildingblocks.
Itisacom-monbelievethatsoftwarequalityincreasesbypatternapplication.
But,theimpactofapatternontopropertiessuchasperformance,securityorenergyconsumptioniswidelyunknown.
Inthispaper,wecomparetheimpactofdesignpatternsontotheenergyconsumptionofmobile(i.
e.
,smartphonebased)applications.
SmallappsfortheAndroidplatformweredevelopedthateitheruseornotuseaspecicpattern.
TheenergyconsumptionoftheseappswasmeasuredbyusingthePowerTutor-App,developedattheUniversityofMichigan.
Theresultsregardingtheselectedpatternsubset(facade,abstractfactory,observer,decorator,prototype,andtemplatemethod)areinteresting.
Especiallythedecoratorpatternshowasignicantnegativeimpactontoenergyconsumption.
Background.
Theresearchpresentedinthispaperisrootedintheresearcheldsofenergy-awarecomput-ingandenergyrequirementascertainmenttechniques.
Manyenergy-awareapproacheseithertrytoreduceen-ergyneedsbysubstitutinghardwareresources[2],orbybalancingenergyrequirementsandinformationquality[3].
In[4]itisillustratedthatasimplesubstitutionoftheresourcescentralprocessingunit,andmemoryhelpstoreducetheamountofenergyrequired.
Theauthorsof[5]showedthatprocessinglessprecisedatarequireslessenergy,andalsopresentasetupformeasuringtheenergyrequirementsofcoreandmemoryofamicrocontrollerbasedsystem,runningsortingalgorithms.
Energymeasurementforsoftwarecaneitherbebasedonhardwareorsoftware-basedapproaches[6].
[7]providesanapproachforgeneratingenergymodelsformobilesystemsbyusingthesmartbatteryinterfaceaccompaniedbymeanstoachieveaccuracy.
ToolssuchastheNokiaEnergyProlerorPowerTutor[8]enabledeveloperstomonitorpowerconsumption.
Thesetoolsarebasedonanunderlyingcostmodelthat,itself,isderivedbyanalyzingaspecicdevice(i.
e.
NokiaS60).
[9]presentsapowermodelingschemeandanimplementationthatallowsne-grainedenergyaccounting.
Insoftwareengineering,apatternisageneralre-peatablesolutiontoacommonlyoccurringproblem[1].
Apatternisanabstracttemplatethatneedstoberenedandadaptedbeforeitcanbeintegratedintothecode.
Patternsfocusondescriptionsthatcommunicatethereasonsfordesigndecisions.
But,littleisknownabouttheirimpactontosystemproperties[10].
[11]examinestheimpactofusingdesignpatternsontoper-formanceandprovideaprocessforpatternselection.
[12]presentsanapproachformappingsoftwaredesigntopowerconsumptionandanalyzehowdesigndecisionsaectanapplication'senergyusage.
[13]analyzessixdesignpatternsandexploretheeectofthemonenergyconsumptionandperformance.
ExperimentsExistingapproacheshaveshownthatpatternusageimpactsenergyconsumptionatleastwhenitcomestoembeddedsystems[13]orC++basedcode[12].
Thegoalofourresearchwasevaluatingtheimpactofpatternsontotheenergyconsumptionofmo-bilesystemsthatuseJava.
Theunderlyinghypothesisbeingthatusingornotusingaspecicpatternwillsignicantlychangeenergyneeds.
Resultscanthenbeusedasastartingpointforfurtherexplorationinordertoidentifywhyandhowthedesignpatternsimpacten-ergyconsumption.
WeselectedasubsetoftheGammapatterns(facade,abstractfactory,observer,decorator,prototype,andtemplatemethod).
Toevaluatetheim-pactofasinglepatternwedevelopedtwo,comparableapplicationsforeachpatternthateitheruseornotusethepattern.
Implementationlooselyfollowedstandardimplementationsavailableintextbooks.
Energycon-sumptionwasthenmeasuredbyusingthePowerTutorApprunningonvariousphones(NexusOne,GalaxySII,Transformer)wherebyexperimentsandmeasurementswerecoordinatedbyaframework.
Figure1:ExperimentalResultsTable1showstheresultsoftherstexperimentseries.
WhilemeasurementsforpatternssuchasFacade,ObserverorTemplateMethodshownodierence,theresultsforthePrototypeandDecoratorshowalargedierenceintimeandenergyneeds(15,2vs.
35.
4and374vs.
873,8).
Thereasonforthegrossdierencemightbethelargeamountofobjectsinstantiationsandmethodcallsofthepattern-basedsystem.
Thissupportsthendingsof[4]thatmemoryconsumptionusingtheheapaswellasthegarbagecollectorareenergy-intensiveoperationsthatalsohaveanegativeimpactontoperformance.
Althoughinteresting,ourresultscanonlybeusedasanindicatorduetoseveralthreatstovalidity.
Implementationandmeasurementmightnotbegeneralizable.
Thiswarrantsfurtherresearch.
SummaryandConclusions.
Inthispaper,wepre-sentedacasestudythatexaminedtheimpactofdesignpatternapplicationontoasystemsenergyconsumption.
Twogroupsofapps,eitherusingornotusingapattern,weredevelopedandmeasured.
TheresultsforadistinctsubsetoftheGammapatternsshowed,thatespeciallythedecoratorpatternhasanegativeimpactontheen-ergyneedsofanapp.
Duetothelowtemporalresolutionofthesoftwaremeasurementmethod,evaluationswithashortruntimeareerror-proneandtheusedsystemsmightnotberepresentable.
However,theinterpretationoftheevaluationresultssupportsourhypothesisandjustiesfurtherresearch.
Usingpatternsisnotalwaysagoodidea.
Theirselectionshouldnotsolelybebasedontofunctionandstructurebutalsoaccordingotherproperties.
Althoughtheresultsofourstudyarenotgeneralizable,theresultsindicate,thatfurtherresearchiswarrantedthatexaminestheimpactofpatternsre-gardingdierentplatformsandapplications.
Resultscanthenbeusedformeta-analysis.
Duringourstudy,wewereabletosupportourhy-pothesesbut,inturn,alsoidentiedissuesthatwarrantfurtherresearch.
First,therobustnessofourapproachregardingthehardwareplatformhastobeevaluated.
Furthermore,itisinterestingtotakeadeeperlookintothecharacteristicsoftheenergyrequirementsofotherpatternsoridioms.
Resultsmightthenbeusedtodeneanti-patternsregardingsoftwareenergyconsumption.
References[1]E.
Gamma,R.
Helm,R.
E.
Johnson,andJ.
M.
Vlissides,"Designpatterns:Abstractionandreuseofobject-orienteddesign,"inECOOP,1993.
[2]C.
BunseandH.
H¨opfner,"Resourcesubstitutionwithcomponents—optimizingenergyconsumption,"inICSOFT'08Proc.
,pp.
28–35,INSTICC,2008.
[3]J.
P.
Sousa,R.
K.
Balan,andD.
G.
etal,"UserGuidanceofResource-AdaptiveSystems,"inICSOFT'08Proc.
,INSTICC,2008.
[4]C.
Bunse,H.
H¨opfner,S.
Roychoudhury,andE.
Man-sour,"Energyecientdatasortingusingstandardsortingalgorithms,"inSoftwareandDataTechnolo-gies,pp.
247–260,Springer,2011.
[5]H.
H¨opfnerandC.
Bunse,"EnergyAwareDataMan-agementonAVRMicroControllerBasedSystems,"ACMSIGSOFTSEN,vol.
35,May2010.
[6]C.
BunseandH.
H¨opfner,"Ocemes:Measuringoverallandcomponent-basedenergydemandsofmobileandembeddedsystems,"inGI-Jahrestagung,2012.
[7]M.
DongandL.
Zhong,"Self-constructivehigh-ratesystemenergymodelingforbattery-poweredmobilesystems,"inProc.
ofthe9thIntl.
conferenceonMobilesystems,applications&services,ACM,2011.
[8]L.
Zhang,B.
Tiwana,andZ.
Q.
etal.
,"AccurateOnlinePowerEstimationandAutomaticBatteryBehaviorBasedPowerModelGenerationforSmartphones,"inCODES/ISSS'10Proc.
,ACM,2010.
[9]A.
Pathak,Y.
C.
Hu,andM.
e.
a.
Zhang,"Fine-grainedpowermodelingforsmartphonesusingsystemcalltracing,"inProc.
ofthe6thconferenceonComputersystems,ACM,2011.
[10]D.
GrossandE.
S.
K.
Yu,"Fromnon-functionalrequirementstodesignthroughpatterns.
,"Requir.
Eng.
,vol.
6,no.
1,pp.
18–36,2001.
[11]N.
Mani,D.
C.
Petriu,andM.
Woodside,"Towardsstudyingtheperformanceeectsofdesignpatternsforserviceorientedarchitecture,"inProc.
ofthe2ndWOSP/SIPEWintntlconferenceonPerformanceen-gineering,ICPE'11,ACM,2011.
[12]C.
Sahin,F.
Cayci,andI.
L.
M.
e.
a.
Gutierrez,"Initialexplorationsondesignpatternenergyusage,"in1stInternationalWorkshoponGreenandSustainableSoftware(GREENS),ACM,2012.
[13]A.
Litke,K.
Zotos,andE.
C.
etal,"Energyconsumptionanalysisofdesignpatterns,"inProc.
WorldAcademyofScience,Engineering&Technology,2005.