its engineering The GREENSOFT Model: A reference model for green and sustainable softwareand Sustainable Computing: Informatics and Systems

ContentslistsavailableatScienceDirect

Sustainable Computing: Informatics and Systems

j ou rna l h o me pa g e : w w w . e l s e v i e r . c o m / l o c a t e / s u s c o m

The GREENSOFT Model: A reference model for green and sustainable software and its engineering

Stefan Naumann

, Markus Dick, Eva Kern, Timo Johann

TrierUniversityofAppliedSciences,Umwelt-CampusBirkenfeld(EnvironmentalCampusBirkenfeld),ISS-InstituteforSoftwareSystems,P.O.Box1380,55761Birkenfeld,Germany^1,2,3

a r t i c l e i n f o

Articlehistory:

Received6May2011 Accepted21June2011

Keywords:

GreenSoftware GreenIT GreenbyIT SoftwareEngineering GREENSOFTReferenceModel

a b s t r a c t

TheresourceandpowerconsumptionofICTisstillincreasing,butalsothebenefitsofICT,e.g.infinding moreefficientsolutionsforenvironmentalproblems.Todate,itisnotclear,whethertheresourceand energysavingsthroughICToverbalancetheresourceandenergyconsumptionbyICT,ornot.Uptonow, manifoldeffortsofGreenITaddresstheenvironmentalaspectsofsustainabilityconsideringcomputer hardware.However,thereisstillalackofmodels,descriptionsorrealizationsintheareaofcomputer softwareandsoftwareprocessmodels.Inourcontribution,wefirstproposedefinitionsoftheterms

“GreenandSustainableSoftware”and“GreenandSustainableSoftwareEngineering”,thenweoutlinea conceptualreferencemodel,theGREENSOFTModel.Thismodelincludesacradle-to-graveproductlife cyclemodelforsoftwareproducts,sustainabilitymetricsandcriteriaforsoftware,softwareengineering extensionsforsustainablysoundsoftwaredesignanddevelopment,aswellasappropriateguidance.

© 2011 Elsevier Inc. All rights reserved.

1. Introduction

Green house gaseffects, climatechange and, asa meansto mitigatethese,SustainableDevelopment(abbr.SD)arethemajor challenges,mankindisfacedwithintheworldoftoday[1].Infor- mationand CommunicationTechnology(abbr. ICT)makesup a considerableconstituentofthesecomplexchallenges.Ontheone hand,ICThasthepotentialtopushSD,ifit isusedtooptimize materialflowsortosubstitutematerialproductswiththeirvir- tualcounterparts,whichreducesenergyandresourceconsumption [2].On theotherhand, itsever-increasingusageinduces rising demandsforenergyandresources[3].Asaneffect,theapproxi- matedenergyconsumptionofU.S.datacentersincreasedfrom28 billionkWhin2000upto61billionkWhin2006[4].Meanwhile, estimationsrosefrom58billionkWhin2000upto123billionkWh in2005onaglobalscale[5].

Previousacademicresearchdiscussedtherelationshipbetween SDandICT.TheseworksfocusontheimpactsofICTonenviron- mentalsustainability[6]oronthebalancebetweenenergyand resourcesavingsbyICTandenergyandresourceconsumptionof ICT[7].Unfortunately,uptonow,noconsenthasbeenfoundon

∗Correspondingauthor.Tel.:+496782171217;fax:+496782171268.

E-mailaddresses:s.naumann@umwelt-campus.de(S.Naumann), m.dick@umwelt-campus.de(M.Dick),e.kern@umwelt-campus.de(E.Kern), t.johann@umwelt-campus.de(T.Johann).

1 http://www.umwelt-campus.de/.

2 http://iss.umwelt-campus.de/.

3 http://www.green-software-engineering.de/.

whetherenergyandresourcesavingsbyICTwillexceeditsenergy andresourceconsumptionornot.

OurcontributionpresentstheGREENSOFTModel,amodelof

“GreenandSustainableSoftware”addressingbothchallenges:the reductionoftheenergyandresourceconsumptioninICT,aswellas theuseofICTtocontributetoSD.Ingeneral,theproposedreference modelcanbeclassifiedintothenewresearchfieldofSustainability Informatics[8].

2. Backgroundandrelatedwork

Untilnow,therearemanypublicationsavailablediscussingthe relationshipbetweenICTandSD.BerkhoutandHertin[9]identified threemainimpactsofICTontheenvironment,whilesummariz- ingliteratureonthetopic.Theseimpactsare:first-,second-,and third-orderimpacts.First-orderimpactsareenvironmentaleffects thatresultfromproductionanduseofICT,i.e.resourceuseand pollutionfrommining,hardwareproduction,powerconsumption duringusage,anddisposalofelectronicequipmentwaste.Second- orderimpactsareeffectsthatresultindirectlyfromusingICT,like energyandresourceconservationbyprocessoptimization(dema- terializationeffects),orresourceconservationbysubstitutionof materialproductswiththeirimmaterialcounterparts(substitution effects).Third-orderimpactsarelongtermindirecteffectsonthe environmentthatresultfromICTusage,likechanginglifestyles thatpromotefastereconomicgrowthand,atworst,outweighthe formerlyachievedsavings(reboundeffects).Theseeffectsdonot appearsequentiallyanddisconnected.Inrealitytheyarenested, which meansthat second-ordereffects canonlyemergeonthe 2210-5379/$–seefrontmatter© 2011 Elsevier Inc. All rights reserved.

doi:10.1016/j.suscom.2011.06.004

basisoffirst-ordereffectsandthird-ordereffectscanonlyappear asramificationsofsecond-ordereffects.

Thesefindings,whichfocusmainlyonenvironmentalaspects, canberefinedtoaddressalsohumanandsocialsustainabilityissues ofICTproductionanduse[6].Here,theeffectsofICTsarecalled

“effectsofICTsupply”(first-ordereffects),“effectsofICTusage”

(second-ordereffects),and“systemiceffectsofICT”(third-order effects).TheysubstantiallydescribethesameeffectsasBerkhout andHertindescribed,butwithoutlimitingthesetoenvironmental issues.

Hilty[10]developedamodelthatcombinesstandardLifeCycle Assessment(abbr.LCA,)[11,12]withtheaforementionedeffects, toshowthepotentialimpactsofICTsonthelifecycleofotherprod- ucts.HeappliedhismodeltofieldsofICTapplications,whichhave beensaidtohaveahighcarbondioxidereductionpotentialandhe identifiesobstaclesthathindertheuseofthesepotentials.

Fuchs[13]discussestherelationshipofICTandSDanddisman- tlesthemythsthatteleworkinghasreducedtheneedtotraveland thatinformationeconomyisnearlyweightlessanddematerialized.

Hiltyetal.[14]showedthatnewversionsofasoftwareproduct oncurrenthardwarearenotnecessarilymoreproductivethanthe olderversionandcanbeevenlessproductive.

Behrendetal.[15]investigatedtheproblemofconflictsthatare exacerbatedbytheextractionofrawmaterialswidelyusedinelec- tronicproductsandcomponents.Relatedtosocialsustainability, theseconflictscanbevaluedassocialsustainabilityrelatedthird- ordereffectsofICTproduction.First-andsecond-orderimpactson socialandhumanacceptabilityofICTproductionandelectronic wastedisposalhavebeeninvestigatedbyBormanandPlank[16], ChanandHo [17], PrakashandManhart [18],and many others [19–22].Theirdiscussionsrangefromwagesbelowthestatutory minimumtoexcessiveworkinghoursandfromforcedovertimeto healthandsafetyrisks.⁴

CoroamaandHilty[7]investigatedthetwopopulardirectionsof ICT:reducingtheenergyconsumptionofICTitselfandreducingthe environmentalimpactsonothersectorsbyICT.Theymakethecase thatthesetwodirectionsshouldbeintegratedinordertoassessnet greenhousegasemissionsinspecificapplicationareasofICT.They arguethatreportsinthisarea(e.g.[23])independentlylookatthese twoaspectswithouttakingintoaccountthatenergysavingsinICT applicationareasmaybeovercompensatedbythegrowthofICT appliancesthatarenecessarytoachievethesesavings.

Mocigemba[24]introducedasustainablecomputingconcept thatcanbeusedtoclassifyandunderstandthedifferentdirections inthisarea.Thisconceptimplementsthreedifferentfociatatime forthemateriallevelofICT(hardware)andfortheinformational levelofICT(software).Thefirstlevelfocusesontheproduct,thesec- ondontheproductionprocess,andthethirdontheconsumption process.Thegivenexamplesmainlydealwithsocialandhuman issuesofcomputeruse.Theyhardlyconsiderneitherenvironmen- talissuesnorthatenergyandresourceconsumptionofICTisrelated tothepotentialofenergyandresourceconservationinducedbyICT.

Abenius[25]relateseffectsofICTproductstotheirlifecycle phasesandidentifiesthat“GreenSoftware”canmitigatefirst-order impactsofICTusage.Sheproposesagroupingof“GreenSoftware”

into“existingtools” and “newinventions”,which are thenfur- therdividedintothecategories“monitorandmeasure”aswellas

“increasingperformance”.

Albertaoetal.[26,27]relatecommonsoftwarequalityaspects likemodifiability,portability,orperformancetotheinterdepen- dentareasofSD,namelyeconomy,society,andenvironment.They alsopresent somemetrics,usuallyusedtomeasurethesequal-

4 Manymorepublicationsinvestigatingtheseandrelatedissuescanbeeasily foundhere:http://ewasteguide.info/biblioarchive.

ityaspects,butlackshowinghowtheresultingvaluesshouldbe interpretedinordertoachievetheintendedeffectsonSD.Besides these,theyproposeasimpleimprovementcyclethatcanbeapplied insoftware developmentprojectsand thathastheobjectiveto improvethesustainabilityofthesoftwareproductfromonerelease tothefollowing.

Arndtetal.[28]discussimplicationsofevolvingreleasesofa widelyusedtextprocessorandrelatethesetoGreenITandSD.As asolutiontocopewithsustainabilityissuesduringsoftwaredesign anddevelopment,theyproposethesocalled“GrandManagement InformationDesign”,which triestotransfertheBauhaus design principlestoimmaterial software products.Dickand Naumann [29]presentedagenericenhancementforsoftwaredevelopment processesthatinstitutionalizestheconsiderationofsustainability issuesduringsoftwaredesignanddevelopment.

Kansaletal.[30]introduced“Joulemeter”,atoolthatestimates theproratapowerconsumptionofvirtualizedserversrunningon onehardwareserver.It usesapowermodelthatleveragesCPU usageand diskIOofeachvirtualserverinordertoestimateits powerconsumption.Thespecificpowerparametersofserverhard- wareare determinedwith powermeters(either aninternal or externalone),whereasthemodelparametersarelearnedinsitu.

ZapicoandTurpeinen[31]introduced“Greenanalytics”,atoolthat visualizestheimpactof websites ontheenvironment byusing datafromGoogleAnalytics.AmselandTomlinson[32]presented

“GreenTracker”, a tool thatestimates the energy consumption of software in order tohelp users tomake informed decisions aboutthesoftwaretheyuse.Bothtoolshavetheobjectivetoraise awarenessofsoftwareinducedenergyconsumptionamongusers aswellassoftwaredevelopers.Naumannetal.[33]presentedthe

“PowerIndicator”(nowcalled“GreenPowerIndicator”),anadd- onforapopularwebbrowser,whichvisualizes,whetherornotthe webserverthathoststhecurrentwebsitespoweredbyrenewable energy.

Capraetal.[34]investigatedtheimpactsofdifferentsoftware systemsonITenergyconsumption. Theystatethatenergy effi- ciencyissuesaremainlyfocusingonhardware.However,inthe rarecasesinwhichtheyfocusonsoftware,itusuallyisonembed- dedsoftware,whereenergyisastrictlylimitedresource.Theyfind thatdifferentsoftwareproductsthatsatisfythesamefunctional requirementsdiffersignificantlyintheirdirectenergyconsump- tion.Furthermore,theyshowedthatimprovingtimeperformance ofasoftwareproductmaynotnecessarilyleadtolowerenergy consumption.

Therearealsosomesupportingtoolsandguidelinesavailablefor softwareengineersthatpresentbestpracticesonhowenergyeffi- ciencyofsoftwarecanbeoptimized,e.g.[35],onhowdatatransfer volumesofwebsitescanbereduced,e.g.[36–38],oronhowsoft- wareartifactscanbeinstrumentalizedandanalyzedinorderto measuretheenergyconsumptioninducedbythem[39,40].

3. WhatisGreenandSustainableSoftware?

Beforepresentingourmodel,itisnecessarytoclarifywhatwe understandby“GreenandSustainableSoftware”and“Greenand SustainableSoftwareEngineering”.Hence,wegivetwodefinitions inthissection.Thesedefinitionsarebasedonthebackgroundof holisticproductlifecyclesinthesenseofLCAora“cradle-to-grave”

approach,thefindingsonthethreedifferentlevelsofimpactsof ICTsonSD,andHilty’sworkonimpactsofservicesofferedbyICTs onthelifecyclesofotherproductsandservices.

In principle,a software product that isattributed as“Green andSustainable”shoulditselfbeassustainableaspossible.This meansthateconomic,societal,andecologicalimpacts,aswellas impactsonhumanbeings thatresultfromtheproduct overits

wholelife cycle, should beas small as possible.Most obvious arethefirst-ordereffects (or:effectsofICTsupply),likeperfor- mancerequirements,networkbandwidth,hardwarerequirements, andproductpackagingthat directlylead toa moreorless spe- cific demand of energy or natural resources. The second-order effects (or:effects of ICTusage) evolve fromusingtheservices offered by ICTson thelife cycle of other products or services.

Today,theservicesofferedbyICTsareusuallyrealizedbysome kindof software.Therefore, software playsa significantrole in thelifecyclesofmanyotherproductsorservices:softwarecan be used tooptimize product design, production processes, the end-of-lifetreatment,ortheusageofotherproductsorservices.

Unfortunately,thesesecond-ordereffectsarenotasobviousasthe first-ordereffects.Evenhardertopredictoranalyzearethird-order effects(or:systemiceffectsofICT),becauseofthemanifoldsys- temicinterdependencies,which requireexperiencedknowledge fromexaminers.Oneexamplearereboundeffectsthatmayoccur, ifaspecificoptimizationfreesusedresources,whichcanbeused toproducemoreproducts,whichthencausesadditionaldemand fortheseresources.Thismayinturnovercompensatetheinitially achievedsavings. First-ordereffects dealwith theterm“Green IT”, second- and third-ordereffects are connected with“Green byIT”.

Ascanbeseenfromthepreviousparagraph,savingenergyor resourcesbyoptimizingICTsisonlyoneaspect.Theequallyimpor- tant aspectcoversenergy and resourcesthat canbe conserved bytheusageofICTsonotherproductsandservices.Seenfroma broaderpointofview,notlimitedtoresourcesorenergy,theques- tionishownegativeimpactsonecology,society,humanbeings, andeconomycanbemitigatedandhowpositiveimpactsonthese canbepromoted.

Theproblemhereisthatthereissoftwarethatdirectlypromotes sustainabilityaspects,likeresourceorenergyefficiency,becauseit isitsintendedpurpose,e.g.softwarethatenablessmartheating, smartlighting,smartlogistics,paperfreeoffices,etc.Inthesecases, itisrelativelyeasytoassesssecond-ordereffectsofthesoftware.

However,thereisalsomultipurposesoftware,likewordprocessors, spreadsheets,orgraphicssoftware.Forthese,itisnearlyimpossi- bletoassesssecond-orthird-orderimpactsthatresultfromusing thesoftwareproduct,becausesoftwaremanufacturersusuallydo notknowforwhichpurposestheirsoftwareproductisused.Hence, asustainablesoftwareproductitselfshouldhavealowimpacton SDandifitisitspurpose,itshouldpromoteSD.Thesebasicrequire- mentsforgreenorsustainablesoftwareareexpressedinDefinition 1.

Definition1. “[GreenandSustainableSoftware]issoftware,whose directandindirectnegativeimpactsoneconomy,society,human beings,andenvironment thatresultfromdevelopment,deploy- ment,andusageofthesoftwareareminimaland/orwhichhasa positiveeffectonsustainabledevelopment”[41].

However,agreenandsustainablesoftwareproductcanonlybe achieved,ifadevelopingorganizationisawareofnegativeandpos- itiveimpactsonSDthatwilllikelybecausedwhenusingit.Inorder toenablethevariousstakeholderstorecognizetheseimpacts,itis necessarytoinstitutionalizetheirassessmentandrecognitionin theappliedsoftwaredevelopmentprocesses.Thismakessustain- abilityissuesmanageableandputssoftwarearchitects,designers, anddevelopersinapositiontooptimizetheirsoftwareproduct accordingly.Additionally,itisnecessarythatthedevelopmentpro- cessitselfisenvironment-friendly.Thesetwoaspectsareexpressed inDefinition2.

Definition2. GreenandSustainableSoftwareEngineeringisthe artofdevelopinggreenandsustainablesoftwarewithagreenand sustainablesoftwareengineeringprocess.Therefore,itistheart

of definingand developing softwareproductsina way,sothat thenegativeandpositiveimpactsonsustainabledevelopmentthat resultand/orareexpectedtoresultfromthesoftwareproductover itswholelife cyclearecontinuouslyassessed,documented,and usedforafurtheroptimizationofthesoftwareproduct[29].

4. TheGREENSOFTModel

4.1. Overviewofthemodel

The GREENSOFT Model is a conceptual reference model for

“GreenandSustainableSoftware”,whichhastheobjectivetosup- port softwaredevelopers,administrators, andsoftware usersin creating,maintaining,and usingsoftwareinamore sustainable way.Themodel(seeFig.1)comprisesaholisticlifecyclemodelfor softwareproducts,sustainabilitycriteriaandmetricsforsoftware products,proceduremodelsfordifferentstakeholders,andrecom- mendationsforaction,aswellastoolsthatsupportstakeholdersin developing,purchasing,supplying,andusingsoftwareinagreen andsustainablemanner.

ThereferencemodelcontainsaLifeCycleofSoftwareProducts.

Thatis,incontrasttotraditionallifecyclesofsoftware,gearedto LifeCycleThinking(abbr.LCT),whichfollowsthemotto:“fromcra- dletograve”.LCThastheobjectivetoassesstheecological,social, human,andeconomiccompatibilityofaproductduringitswhole lifecycle.Itbeginswiththeearlystagesofproductdevelopment andendswiththeproduct’sdisposalandrecycling.Thefindings gainedfromtheseassessmentscanthenbeusedforabalanced optimizationoftheproductorforcomparinga productwithits competitors[42].

ThesecondpartoftheGREENSOFTModeliscalledSustainability CriteriaandMetrics.Itcoverscommonmetricsandcriteriaforthe measurementofsoftwarequality[43]anditallowsaclassification ofcriteriaandmetricsforevaluatingasoftwareproduct’ssustain- ability.Appropriatecriteriaandmetricsmaycomprisemodelsfor themeasurementofsoftwarequality,proceduremodelsforsoft- waredevelopment,aswellasmethodsborrowedfromLCA[11,12].

Here,wedistinguishdirectcriteriaandmetrics(relatedtofirst- ordereffects)fromthosewhichindirectlyconcernsustainability (relatedtosecond-andthird-ordereffects).

ThemodelcomponentProcedureModelsmakesitpossibleto classifyproceduremodelsthatcoveracquisitionanddevelopment ofsoftware,maintenanceofITsystems,andusersupport.Asan example,weproposedagenericextensionforambiguoussoftware developmentprocessesthatenablesthesystematicconsideration ofsustainabilityaspectsduringsoftwaredevelopment[29].

Thelast component ofthemodel contains Recommendations andTools.Thesesupportstakeholderswithdifferentprofessional skilllevelsinapplyinggreenorsustainabletechniquesingeneral, when developing,purchasing, administrating,orusingsoftware products.Possiblerolesaresoftwaredevelopers,acquirersofsoft- ware, administrators, as well asprofessional and privateusers [38,44].

4.2. TheLifeCycleofSoftwareProducts

TheLifeCycleofSoftwareProducts(seeFig.2),includedinthe GREENSOFTmodel,isaLCTinspiredproductlifecyclethatcanalso beattributedwith“fromcradle-to-grave”.Itsobjectiveistoenable stakeholderstoassessimpactsonSDaccordingtothethreediffer- entlevelsofimpacts,asdiscussedinWhatisGreenandSustainable Software?section.

Whenworkingwiththeselevels,stakeholdersshouldbeaware ofholisticallyconsideringalllevels,because:“Naturallyitiseasier tofocusonfirstorderimpactsastheyareimmediateandobvious.

Fig.1.TheGREENSOFTModel,areferencemodelfor“GreenandSustainableSoftware”.

Fig.2.Cradle-to-graveinspiredproductlifecycleforsoftwareproducts,attributedtoexemplarysustainabilityrelevanteffectsofICTs.

Howeverthethirdorderimpactsmightbethemostthreatening ones,andthemostdifficultonestoapproach”[25].Though,second- andthird-orderimpactsalsohavehugepotentialstopromoteSD, thesestillhavetoberecovered[3,23,45].

4.2.1. Thedevelopmentphase

TheDevelopmentPhaseaccountsforimpactsonSDthatdirectly resultfromactivitiesinvolvedinsoftwaredevelopment,aswell asindirectlyinvolvedactivities,e.g.prorataimpactsofcommon corporatedepartments.Environmentalimpactstobeconsidered include,forexample,electricalenergythatisnecessarytopower theworkstationsofsoftwaredevelopersandotheremployees,elec- tricalenergyandnaturalresourcesthatarenecessarytooperatethe ITinfrastructure(e.g.networkingdevices,servers,andstorages), energythatisnecessaryforheatingandairconditioning,electrical energythatisnecessaryforofficeslighting,orenergyfortrans- portationpurposeslikelongdistancebusinesstripsformeetings withcustomersandthedevelopmentteamandeventheemploy- ees’dailywaytowork.Socialimpactscanbeworkingconditions andpaymentofoffshoreworkers(e.g.developers,typesetters), which have consequences for the workers and their families.

Someoftheseimpactscanbemitigatedbyintroducingtelework- ingandteleconferencing,orbyreplacingmaterialproductswith adequate immaterial substitutes (second-order effects). This in turnmayinduce,e.g.changesinorganizations,softwaredevelop- mentmethods,orlifestyles(third-ordereffects).Thedevelopment phase alsoaccounts for impacts fromsoftware maintenance in thesenseofbug solving,becausethis is alsoa softwaredevel- opmentactivityandthereforeshouldbelongtothedevelopment phase.

4.2.2. Thedistributionphaseandthedisposalphase

TheDistributionPhaseaccountsforimpactsonSDthatresult fromdistributingthesoftwareproduct.Thisincludesenvironmen- talimpacts,e.g.ofprintedmanuals(typeofpaperandink),chosen meansoftransport,typeanddesignoftheretailandtransportpack- aging(e.g.plastic,polyurethanefoam,biodegradablematerial),or datamedium(e.g.CD/DVD,USBmemorystick).Furthermore,ifthe softwareproductisofferedasadownload,whichiscommontoday, thenitsdownloadsizeshouldbeconsidered,aswellastheelectri- calenergyandmaterialresourcesthatarenecessarytooperatethe requiredITinfrastructure.

The Disposal Phase accounts for impacts on SD that result fromdisposalandrecyclingoftheaforementionedmaterialsub products.

4.2.3. Theusagephase

TheUsagePhaseconsidersimpactsthatresultfromdeploying, using,andmaintainingthesoftwareproduct.

Here,maintainingmeansthatadministratorsareinchargeof installedsoftwareandsupportusersintheirorganization.Thus, maintainingincludes,e.g.theinstallationofsoftwarepatchesor updates,theconfigurationofsoftwareandcomputersystems,and thetrainingofemployeesinregardstopropersoftwareusage.As aneffect,properlytrainedusersmightneedlesstimetocomplete tasks(whichresultsin lessenergy consumption),configurethe softwaresysteminawaythatitconsumeslesspower,orjustswitch theircomputertosuspendmodewhentheyleavetheirworkplace.

Beside these effects, software usage has several first-order effectsregardingenvironmentalsustainability.

In order todeliverits offeredservices,a computer program requiresprocessingtime,whichinturnconsumeselectricenergy.

Thismayalsorequiretheconsumptionofservicesofferedbyother servers(consider,e.g.DataBaseManagementSystems,Enterprise ResourcePlanningsystems,orsimplytheWWWservice),which causesadditionalpowerconsumption.

Inaddition,theupdatestrategyofasoftwareproduct(e.g.size andfrequencyofupdates)influencesdatatransfer,processing,and ITinfrastructure, which are necessarytodeliverupdates. Com- bined,thesecausefurtherpowerandresourceconsumption.

State-of-the-art software systemsusually require up-to-date andmorepowerfulhardwarethanoldersoftwaresystemsorpre- viousversions.Asaresult,thiscauseshardwarereplacementsin organizationsaswellasathome,whenanewsoftwareproduct isintroduced.Ontheonehand,newhardwareistypicallymore powerefficientthanolderhardware,butontheotherhandithas tobetaken intoaccount that theproduction ofthe newhard- wareandthedisposaloftheoldhardwarecausesvastamountsof resourceandenergyconsumption[46].Miningthenecessaryores, e.g.indevelopingcountries,wheresocialandenvironmentalstan- dardsareverylow,leadstoconsiderablesocialandenvironmental impacts,whichsometimesevenculminateinarmedconflicts[15].

Therearealsoreportsaboutoldandevennon-functionalhardware thatisexportedfromindustrialcountriestodevelopingcountries, whereitisreusedbutmoreoftenrecycledunderdoubtfulcircum- stancesinsocalledbackyardfacilitiesorjustdepositedonwaste disposalsites,causingdamagetotheenvironment andpeople’s health[46].

The second- and third-order effects onSD that result from the usage phase, depend on the purpose of the software productandwerebrieflydiscussedinWhatisGreenandSustainable Software?section(smart-technologies,dematerialization).

4.2.4. Thedeactivationphase

Ifasoftwareproductistakenoutofservice,itismostlynecessary toconverttheavailabledatatoaformatthatcanbeprocessedbythe succeedingsoftwareproduct,ortomakeitaccessibleinsomeother ways.Ifthedatacannotbeconvertedeasily,e.g.becauseitisstored inaproprietarydataformat,thismayhaveanimpactoneconomic sustainabilityofanorganization.Inthisphase,eventhebackupsize ofdatamatters,e.g.iflegalregulationsrequirelong-termstorage ofdata.

4.3. Sustainabilitycriteriaandmetrics

Ourmodelhastheabilitytorepresentthreecategoriesofsus- tainability criteria and metrics for software products: Common QualityCriteriaandMetrics,DirectlyRelatedCriteriaandMetrics,and IndirectlyRelatedCriteriaandMetrics.Thefirstrelatestocommon qualitycriteriaforsoftware,whicharewellknownfrom,e.g.[43].

Thesecondcomprisescriteriaandmetricsthatrelatetofirst-order effects(effectsofICTsupply).Thelastincludescriteriaandmetrics gearedtowardssecond-order(effectsofICTuse)andthird-order effects (systemiceffectsof ICT).Allcriteriashouldalsobeclas- sifiedaccordingtothephasesofourproposedsoftwareproduct lifecycle.Additionally,itisalsonecessarytoclassifycriteriaand metricsaccordingtothetypeofsoftware.

Albertaoetal.[26,27]interpretcommonsoftwarequalityprop- erties and associated metrics on the background of SD. They classifythequalitypropertiesintodevelopment,usage,andpro- cessrelatedproperties.However,theydonotclassifytheeffects accordingtothethreetiermodelofeffectsofICTsonSD.Notwith- standingthis,theirfindingscanbesubsumedbytheGREENSOFT Model.

Thequestions in thelifecycleof theGREENSOFT Modelare not:inwhichphasesaremetricsapplied,orinwhichphasesare measurestaken, inorder toimprove thecorrespondingquality properties?Rather,thequestionis:inwhichlifecyclephasecan therelatedeffectsbeobserved?

Hence, the quality properties “Modifiability” and “Reusabil- ity”takeeffectintheDevelopmentPhase,whereastheproperties

“Portability”, “Supportability”, “Performance”, “Dependability”,

“Usability”,and“Accessibility”takeeffectintheUsagePhase.The processrelatedproperties“Predictability”(“Theteam’sabilityto accuratelyestimateeffortandcostupfront”[26])and“Efficiency”

(“Theoverheadofproductionprocessesoverthebottomlinevalue perceivedbythecustomer”[26])alsotakeeffectintheDevelop- mentPhase,aswellasthe“Project’sFootprint”(“Naturalresources and environmentalimpact used during software development”

[26]).

Theproperty“Portability”isinterpretedbyAlbertaoetal.on thebackgroundofhardwareobsolescence,whichmeansthatthe lifetimeofhardwareshouldbeprolongedtotheendofitsuseful lifetime,insteadofcausingitsearlyreplacementduetohardware requirementsimposedbyasoftwareproduct.AccordingtoHilty [46]basedonaLCAstudybyEugsteretal.[47],desktopPCsshould notbeusedforlessthanapprox.5years.After5yearsofuse,the ecologicalimpactsresultingmainlyfromtheenergydemandofthe usagephaseoutweighthoseoftheproductionphase.Thesefindings arealsosupportedbyarecentLCAstudybyamanufacturerofICTs [48].Thisisdifferentforservers,becauseaserverin24/7operation modereachesthepointofequilibriumearlier[46],accordingto thefigurespresentedby[49]afterapprox.1yearofuse.Duetothe factthattherewillbehigherratesofrenewableenergyandmore energyefficienthardwareinthefuture,itwilltakeevenlongeruntil theenvironmentalimpactsoftheusagephaseoutperformtheseof theproductionphase.Forthatreason,softwareinducedhardware obsolescenceisofparticularimportance.Hence,hardwareobsoles- cenceshouldbeagenuinequalitypropertyofgreenandsustainable software,whichbelongstotheDirectlyRelatedCriteriaandMetrics modelpart.

Anotherdirectlyrelatedqualitypropertyisenergyefficiency.

Thisisnotthesameasruntimeefficiencyorperformance,because itsgoalistooptimizeenergy consumptionin relationtodeliv- eredserviceitems.Forasoftwarerunningonasinglecomputer, energyefficiencyandperformancemaybecloselyrelated,butthere maybegreaterdifferences fordistributedsystems. Inthis area, consider,e.g. service level agreementsor performance require- mentsthat necessitateadditionalservers tohandle peak loads.

Here,it maybepossibletoincreaseenergy efficiencybyrelax- ingperformancerequirementsforpeakloads.Anotherpossibility maybetherelaxationofrequiredservicequalitydowntoalevel thatisstillacceptableforusers.Dependingontheareaofapplica- tion,itmaybepossibletouseapproximationsinsteadofaccurate calculations,whichmayrequirelessprocessing.Consider,e.g. a search engine that delivers a huge amountof accurate search resultscomparedtoonethatensures thisaccuracyonlyforthe topmostresults that areused most widelyby the users but is less accurateregarding thefollowing results. Another example aremathematicalcalculations,whichmaybeonlyasaccurateas theyareneededtobeacceptableforthepurposeofthesoftware application.

Inadditiontothesesoftwareartifactsanddevelopmentprocess relatedproperties,therearealsoproperties,whichbelongtothe developingorganization.Organizationsthatdevelopgreenandsus- tainablesoftwareshouldcommitthemselvestoenvironmentaland socialresponsibility,expressed,e.g. inenvironmentaland social responsibilitystatements,theircommitmenttointernationallabor standards[50],ortheapplicationofenvironmentalmanagement systems[51].Thesecommitmentsshouldalsocoverenvironmen- talandsocialstandardsthroughouttheentiresupplychainofall productsandservices,whicharenecessarytoproduce,advertise, distribute,anddispose/recyclethesoftwareproductorpartsofit.

IndirectlyRelatedCriteriaandMetricsforgreenandsustainable softwareaddresssecond-andthird-ordereffectsinducedbyasoft- wareproduct.Usingasoftwareproductmayachieveenergyand resourcesavingsinotherbranchesorusageareas.Althoughitis hardtoidentifythesesecond-andthird-ordereffects,theremaybe

effectsthatcanbemeasuredandapproximatedanyway,i.e.when thesoftwareproducthasaspecialpurposelikesmartheatingand air-conditioningofbuildings.Evenhardertorecognizearesocial effectsconnectedwithasoftwareproduct.Becausetheseeffects specificallydependonthesoftware product’spurposeand area ofapplication,itisnotpossibletodefineappropriatemetricsor criteriatorecognizetheseingeneral.Examplesarehowsocialnet- workschangethewaytocommunicateandtointeract,andtheway productionchainsarealtered,sinceICTsalmostallowworld-wide productioninrealtime.

4.4. Proceduremodels

Starting from our definitions, Green and Sustainable Soft- wareEngineeringproduces Green andSustainableSoftware inan environmental-friendlyandsustainableway.Consequently,dur- ingtheengineeringprocess,thewholelifecycleoftheengineered softwareproducthastobetakenintoaccount,aswellasthecir- cumstancesunderwhichitwillbeproduced.

4.4.1. Sub-proceduremodel“Develop”

Inthissection,wepresentanexampleofasoftwaredevelop- mentprocessthatfitsintothecategoryDevelopoftheprocedure modelpartoftheGREENSOFTModel.

This example [29] (see Fig. 3) does not implement a com- pletesoftwaredevelopmentprocess.Instead,itproposesseveral enhancementsforarbitrarysoftwaredevelopmentprocessesthat enablestakeholderstorecognizeimpacts,whichresultfrom“pro- ducing”thesoftwareproduct,andimpacts,whichresultfromusing thesoftwareproduct.Theproposedenhancementsare:Sustainabil- ityReviews &Previews,ProcessAssessment,SustainabilityJournal, andthesocalledSustainabilityRetrospective[29].Inprinciple,the enhancementsformacontinuousimprovementcyclethat deals withsustainability issues. Process Assessment helps tooptimize thesustainabilityofthe“production”process,whereasSustainabil- ityReviews&Previewshelptooptimizethesustainabilityofthe evolvingsoftwareproduct.BotheffortsarecombinedbytheSus- tainabilityRetrospective,sothatfinallyimpactsoverthewholelife cycleofthesoftwareproductarecovered.

SustainabilityReviews&Previewstakealookattheworkdone, assessoutcomes accordingtosustainabilityissues, and develop measures,whicharerealizeduntilthenextSustainabilityReview

&Previewinordertooptimizethesustainabilityofthesoftware productunderdevelopment.Thesereviewstakesoftwareaspects, likerequirements,architecture,orcodingintoaccountthathave impacts on sustainability. In Sustainability Reviews & Previews, whichtaketheroleofaformativeevaluation,notonlythesesoft- wareaspectsaretakenintoaccount,butalsoimpactsthatresult fromthedevelopmentprocessitself[29].

The Process Assessment activity continuously monitors the developmentprocess.Therefore,differentdatafromthedevelop- mentprocessisgatheredinordertoassessitsimpactsonSD,and toidentifyfactorsthatshouldbeoptimizedinordertoimprovethe process.ThisdatacanalsobeusedasabasistoperformaLCAof thesoftwareproduct[29].

TheSustainabilityRetrospectivesumsupthedatacollectedby SustainabilityReviews&PreviewsandProcessAssessment,assesses theoverall impactonSDofthesoftwareproduct,and looksfor waystoimprovethesustainabilityofupcomingsoftwareprojects andtheirsoftwareproducts.FurtheroutcomesoftheSustainability Retrospectiveare,e.g.assessmentsandgroupreflectionsofimpacts onSDofthedevelopedsoftwareproductandthedevelopmentpro- cess,decisionsforfutureprojects,lessonslearned,orbestpractices regardingsustainabilityissuesofsoftwareproductsanddevelop- mentprocesses[29].

Fig.3.Exampleforenhancingsoftwaredevelopmentprocessesthatfitsintotheproceduremodelpart“Develop”[29].

Itshouldbementioned,thatthisproceduremodelsetsonlyan organizationalframeworkthathelpsmanagingsustainabilityofa softwareproduct.Thus,thereis noguaranteethat theresulting softwareproductsaremoresustainablethantheywouldhavebeen withoutapplyingthisprocess.

Besidesreflectingtheproposedlifecycleofasoftwareproduct, therearefurthermethodsthatsupportsoftwarearchitects,design- ers,anddevelopersinproducinggreenandsustainablesoftware applications.Ontheonehand,therearetoolsthatautomatically calculatesoftwaremetricsfromsourcecodeorcompiledartifacts.

Aswasshownabove,thesemetricsandtheircorrespondingqual- itypropertiescanalsobeinterpretedregardingsustainability.On theotherhand,producingecologicallysound,resourceandenergy efficientsoftwareisalsoanissuenowadays.Forthispurposeearly processingtimeestimations,energyconsumptionestimations,and energyconsumptionmeasurementsmaybeappropriate.

Earlyprocessingtimeestimationscanbeobtainedfromasoft- wareperformance estimationmethod,introducedbySmithand Williams[52].Thismethodestimatesperformancevaluesalready inearlydesignstagesofasoftwareproduct.Itstartswithroughesti- mationsbasedonearlyUMLsequencediagrams[53]andrefines these estimations as theUML model evolves. This resultsin a socalledsoftwareexecutionmodel.Thesystemexecutionmodel modelsthetargethardwareplatformwithitslimitedresources(e.g.

CPUcores,netIO,diskIO).Thesoftwareexecutionmodelcanbe appliedtothesystemexecutionmodelwithadiscreteeventsimu- latorinordertogetperformanceestimationsforrealisticworkloads [52].Thisapproachhelpstodesignsoftwarearchitecturesthatper- formwellonaspecifichardwareplatform,andithelpstoidentify designflawsthatreduceruntimeefficiency.Currently,thereisno softwaretoolavailablethatallowstheestimationofenergycon- sumptionintheseearlydesign stages.Sowerecommendusing the software performance approach as an indicator of energy efficiency.

Inasoftwareproject,whenthefirstdeployablesoftwarearti- factsareavailable, itis easilypossibletomeasure theirenergy consumption,eitherbymeasuringtheenergyconsumptiondirectly [39]orbyusingperformancemonitorcountersofmodernproces- sorsasindicators[40].Thesemeasurementscomplementtheearly estimationsandcanprovidefurtherindicationsonsoftwarecom- ponentsthatinducehighenergyconsumptionandshouldtherefore beoptimizedwithpriority.

Basedonthesesoftwarearchitectureandsoftwaredevelopment centricmeasures,softwaredevelopingorganizationsshouldalso estimatethetotalenergyandresourcedemandthatisexpected

accordingtotheirprojectednumberofinstallationsorsalesfig- uresofaspecificproductandassociatedestimatedusagescenarios.

Theseestimationsshouldnotonlyincludefirst-orderimpactsofthe usagephase,butalsofirst-orderimpactsoftheotherphases,espe- ciallyofthedistributionphase.Asafurtherstepanddependingon thetypeofsoftware,itmayalsobepossibletoestimatesecond- andthird-orderimpactsrespectively.Thesecanbeusedtosub- stantiatethenecessitytoimprovethesustainabilityofasoftware productfromabroaderpointofview.

4.4.2. Sub-proceduremodel“Purchase”

Ouroutlineofanexemplaryproceduremodelthatfitsintothe Purchasecategoryfocusesmainlyongovernmentalorganizations andlargeenterprisesthatusestructuredtenderingproceduresin largescaleprocurementprojects.Especiallythelargemarketpower ofgovernmentalorganizationsshouldnotbeunderestimatedand canbeusedtopursuesustainabilitygoalsaccordingtonational(e.g.

[54])andinternationalagreements(e.g.[55,56]).

Atypicalprocurementprocesshasthefollowingsteps:define subjectmatter,definerequirements,selectbidders,evaluatebids, andconcludecontract[51,57].Duetolegalreasons,itisnecessary thatsustainabilityissues,i.e.ecologicalandsocialrequirements, areclearlystatedinthetender’ssubjectmatter,inspecifications andincontractperformanceclauses[57].Furthermore,theentire productlife cycleofthesoftwareproduct shouldbeaddressed, aswellastheentiresupplychain.Ourproposalforasustainable softwareprocurementprocesscanbedividedintotwofields:the procurementofcustomsoftwareproductsandtheprocurementof standardsoftwareproducts.

Inbothscenarios,biddersshould,inprinciple,beabletodeliver therequestedproductintherequiredquality.Hence,biddersmay bepreselectedwithappropriatecriteria,likethecompany’ssocial andenvironmentalresponsibility(e.g.expressedinenvironmental andsocialresponsibilitystatements),theircommitmenttointer- nationallaborstandards[50]ortheapplicationofenvironmental managementsystems[51].

For custom software products, non-functional requirements likeenergyefficiencyorrequirementsaddressingmitigationofIT infrastructureobsolescence,whichmaybeinducedbythetendered softwareproduct,canbedefined.However,itisnecessarytopro- videapplicablemeasurementmethodsandacceptablemaximum measurementvaluesforbidders.Fromapracticalpointofview,this isonlypossibleifacomparablesoftwareproductisavailablethat canbeusedasreference,e.g.ifalegacysoftwaresystemisreplaced witha newone. Ifsuchvalues cannot beprovided, purchasers

candefinecontractperformanceclausesthatpledgethecontrac- tortoestablishmeasures,whichencourageasustainablesoftware product.Thisincludestheapplication ofa sustainablesoftware developmentprocess,environmentmanagementsystems,orthe assertionofsocialstandardsalongtheentiresupplychain(includes alsosubcontractors).

Forstandardsoftwareproducts,energyefficiencyandhardware obsolescencecriteriacanbeeitherusedastechnicalrequirements orasawardcriteria.Usually,theirlimitvaluesandweightingin theselection proceduremust be documented in the tendering documents,which meansthatthesecannotbealteredafterthe tenderingprocedurehasstarted.

Purchasingsoftwareorpurchasingtheappropriatehardware forasoftwareproductisalsoofhighrelevanceforhomeusers.

Itis clear,thathomeusers,as wellas purchasersofmicro and smallenterprises,requireinformationonsustainabilityissuesthat canbeobtainedeasily.Thismaybeaccomplishedbyprintingthe accordinginformationonproductboxesorproductsheets.Inthis way,thecustomerscanmakeinformeddecisionsonwhichprod- uctfitstheirneedsbest.Hence,theremaybealabelforsoftware productssimilartotheENERGYSTAR^®5thatindicateswhethera softwareproductisenergyefficientormeetscertainsustainability requirementsinthefuture.

4.4.3. Sub-proceduremodel“Administrate”

In our model, “administrate” means making software avail- ablebyinstalling,configuring,andmaintainingit.Thisalsocovers educatingandtrainingusers,whoworkwiththesoftwareinan organization.

Configuringsoftwareproductscoversonlyconfigurationfrom anorganizationalpointofview,herecalled“macro-configuration”.

Configurationsthat canbe donebyusers withouttheneed for specialsystempermissions,e.g. installingadd-onsforbrowsers, configuringawordprocessorviaitspreferencesdialog,etc.,are covered as “micro-configuration” by the sub-procedure model Use. Hence, the macro-configuration covers system configura- tion and maintenance of desktop PCs, thin-clients, as well as serversandvirtualizeddatacenters.Especially,thevariouspub- licationsondatacenterenergyefficiencyfitintothismodelpart (e.g.[58–60]).

Aminimalisticproceduremodelshouldimplementacontin- uous improvement cycle, which follows the plan-do-check-act paradigm.Here,i.e. energyefficiency,energy consumption, and resource consumption should be checkedregularly in order to improvethesewithappropriatemeasures.Thisisnotmerelylim- ited to data center operations, but rather includes networking infrastructure,desktopcomputers,installedsoftware,andusers, servedbyanITservicedivisionofanorganization.Especiallyin organizationsthat implement service desks in linewith theIT InfrastructureLibrary[61]inordertosupporttheirITusers,the missionofservicedesksalsoistoadviseusersproactivelyonsoft- wareconfigurationandusageregardingsustainabilityissueslike energyandresourceconservation.

4.4.4. Sub-proceduremodel“Use”

Proceduresforusers,eitherprofessionalorhomeusers,related togreen and sustainablesoftware, cannot be putinto a static workfloworproceduremodel,becausethesearemainlyad-hoc procedures.Nevertheless, similarlytothesub-procedure model componentAdministrate,acontinuousimprovementcyclecanbe appliedintuitively.Inthiscontext,usersshouldreflectoneachof

5 ENERGYSTAR^® is a registered trademark owned by United States Envi- ronmentalProtectionAgency(EPA);http://www.energystar.gov/;http://www.eu- energystar.org/(2011-03-15).

theiractionsthat mayhave negativeimpactsonSDinorderto searchforappropriateguidancethathelpsmitigatingtheseneg- ativeimpacts.

Asalready mentioned in Sub-procedure model“Administrate”

section,usersdosomekindof“micro-configuration”likeinstalling add-onsinwebbrowsers,configuringthewordprocessor,etc.and theyusesoftwareproducts.However,becauseoftheabsenceofa moreformalizedprocessandthead-hoccharacteroftheactions, proceduresaremoreorlessrepresenteddirectlyasrecommenda- tionsforactionlikeguidelinesorchecklists.

4.5. Recommendationsandtools

RecommendationsandToolsaddressstakeholderswithdiffer- entroles.GeneralrolesconsidereddistinctlybytheGREENSOFT Modelare:Developer,Purchaser,Administrator,andUser.However, theremaybemorespecializedroles,likeRequirementsEngineer, SoftwareArchitect,WebAdministrator,orApplicationDeployer.In principle,thesecanbesubsumedbythegeneralrolesmentioned before.

RecommendationsandToolssupportstakeholderswithdiffer- entskilllevelsinapplyinggreenandsustainabletechniqueswhen developing, administrating or using software products.Recom- mendationscanbeguidelines,checklists,bestpracticeexamples, implementationreports,etc.Toolscanbesoftwaretools,butalso anyothertool,likepaper-baseddatacollectionsheets.Thereare plenty of recommendations available, e.g. on the Internet, but unfortunatelythesearehardtofind.Hence,aspecializedInternet searchengineorknowledgebasewouldmakeiteasiertofindthem [44].

Examplerecommendationsfordevelopersdeal withwebsite optimization [38,62], provide guidance on how to design and develop energy efficientsoftware [39,40,63,64], orgive general hintsonhowtodesignanddevelopresourceefficientsoftware[65].

Availabletoolsfocusonenergyorresourceefficiencyofsoftware orservices[39,66].Foradministrators,thereareguidelinesavail- ablethatdescribebestpracticesforenergyefficientdatacenters [59,60,67,68].Someoftheseguidelinesareaccompaniedbyappro- priatetools[69].Forpurchasers,thereareguidelinesavailablethat describesustainableprocurementofITequipment,inaccordance withtenderingregulations[51,57].Forusers,thereareguidelines andtoolsavailablethatsupportusersinmakinginformeddeci- sionsonhardware,software,andservicestheyuseorplantouse [32,33,70,71].Othertoolslikegrano.la⁶directlyhelpsavingenergy byreconfiguringthepowermanagementofthecomputersystem accordingtocurrentperformancedemands.

5. Discussionandconclusion

Themainobjective ofourproposedGREENSOFTModel isto structureconcepts,strategies,activities,andprocessesofGreenIT andherebyespeciallyofgreen andsustainablesoftwareandits engineering.Ourmodelactsasareferencemodel,whichhelpsto organizeandclassifyresearchresults,actions,frameworks,process models,etc.Additionally,theGREENSOFTModelinvolvesaninher- entroadmap and depicts which direction future developments regardingtheinterconnectionofICTandSDcantake.Themodel alsosuggestshow“conventional”modelsofICT,e.g.standardsoft- wareprocess modelsor tenderingprocedures, can beenriched regardingsustainability.

Consequently,atfirstwe definedwhatGreenandSustainable Softwareisand whatGreenandSustainableSoftwareEngineering

6http://www.grano.la/(2011-03-23).

means.Startingwiththesedefinitionswedescribedinourcontri- butionthefourmainpartsoftheGREENSOFTModel:(1)thelife cycleofasoftwareproductregardingacradle-to-graveapproach and(2)criteriaandmetricsfordirectandindirecteffectsofsoft- wareonSD.Furthermore,we(3)includedproceduremodelsfor developing, purchasing, operating and using sustainably sound softwareinasustainablysoundwayand(4),inordertoputthe modelintopractice,theGREENSOFTModelcontainsaframework forrecommendationsforactionandtools.

We showedthat a softwareproduct isnot merely madeup ofsoftwareartifacts,butratherofmanyotherproductsandser- vices that are involved in a software products’ life cycle. All theseproductsand serviceshave plentyof impacts onSD that must be considered in order to figure out if a software prod- uct is green or sustainable. Furthermore, using the software productleadstoeffectsonSD.Here,especiallysecond-andthird- order effects can either exceed or even out-Herod first-order effects.

OurGREENSOFTModeltakesthoseeffects,thelifecycleofsoft- ware,different userroles, and different activitiesinto account.

Also aspects like “Green IT” vs. “Green by IT” are integrated.

Differentapproaches regarding Green IT,SustainableIT,ICT for Sustainability,etc.canbestructuredandclassifiedintothemodel.

Furthermore,themodelcanbeadoptedinordertounfoldsoftware relatedGreenITactivitiesinbusinessandscience.

However,ineverydecisionregardingGreenIT,onemustbear inmindthat everyadditional softwareproductconsumesaddi- tionalenergy.Consequently,theseeffectsmustbecomparedwith thebenefitsusingthisspecialsoftwareproduct[45]resp.[46].Our GREENSOFTModelhelpstofindacomprehensivesolution.

Acknowledgements

Thispaperevolvedfromtheresearchanddevelopmentproject GreenSoftwareEngineering(GREENSOFT),whichissponsoredby theGerman Federal Ministry of Education and Research under reference17N1209. Thecontentsof this documentarethesole responsibilityoftheauthorsandcanundernocircumstancesbe regardedasreflectingthepositionoftheGermanFederalMinistry ofEducationandResearch.

References

[1] UnitedNationsGeneralAssembly,ReportoftheWorldCommissiononEnvi- ronmentandDevelopment,OurCommonFuture,UNDocumentNo.A/43/427 English, New York, 1987, http://daccess-ods.un.org/TMP/4738853.html (accessed2011-03-09).

[2]K.Fichter,SustainablebusinessstrategiesintheInterneteconomy,in:L.M.

Hilty(Ed.),SustainabilityintheInformationSociety,Metropolis-Verlag,Mar- burg,2001.

[3]L.Erdmann,L.M.Hilty,J.Goodman,P.Arnfalk,TheFutureImpactofICTson EnvironmentalSustainability,2004,http://ftp.jrc.es/EURdoc/eur21384en.pdf (accessed2010-04-26).

[4]U.S.EPA,ReporttoCongressonServerandDataCenterEnergyEfficiencyPublic Law109–431,2007,http://hightech.lbl.gov/documents/DATACENTERS/EPA- datacenters.pdf(accessed2010-10-15).

[5]J.G. Koomey,Estimatingtotal PowerConsumptionby ServersintheU.S.

andtheWorld,FinalReport,February15,2007,2007,http://www.greenbiz.

com/sites/default/files/document/CustomO16C45F77412.pdf(accessed2010- 04-26).

[6]W.Göhring,Thememorandum“SustainableInformationSociety”,in:P.Minier, A.Susini(Eds.),Sh@ring,Proceedingsofthe18thInternationalConference

“InformaticsforEnvironmentalProtection”,EnviroInfo2004,21–23October, 2004,Geneva,CERN,Geneva(Switzerland),ÉditionsduTricorne,Genève,2004, pp.278–286.

[7]V.Coroama,L.M.Hilty,Energyconsumedvs.energysavedbyICT–acloser look,in:V.Wohlgemuth,B.Page,K.Voigt(Eds.),EnvironmentalInformatics andIndustrialEnvironmentalProtection:Concepts,MethodsandTools,Envi- roInfo2009;23rdInternationalConferenceonInformaticsforEnvironmental Protection;Proceedingsofthe23rdInternationalConferenceEnvironmental Informatics– InformaticsforEnvironmentalProtection,SustainableDevelop- mentandRiskManagement,September09-11,2009,HTWBerlin,University

ofAppliedSciences,Germany,vol.2:Workshops,vol.2,Shaker,Aachen,2009, pp.353–361.

[8]S.Naumann,Sustainabilityinformatics–anewsubfieldofappliedinformat- ics?in:A.Möller,B.Page,M.Schreiber(Eds.),EnvironmentalInformaticsand IndustrialEcology;22ndInternationalConferenceonInformaticsforEnviron- mentalProtection;EnviroInfo2008;Proceedingsofthe22ndInternational ConferenceEnvironmentalInformatics–InformaticsforEnvironmentalProtec- tion,SustainableDevelopmentandRiskManagement,September10–12,2008, LeuphanaUniversityLueneburg,Germany,Shaker,Aachen,2008,pp.384–389.

[9] F.Berkhout,J.Hertin,ImpactsofInformationandCommunicationTechnolo- giesonEnvironmentalSustainability:SpeculationsandEvidence,Reporttothe OECD,2001,http://www.oecd.org/dataoecd/4/6/1897156.pdf(accessed2011- 03-02).

[10] L.M.Hilty,CO2reductionwithICT:prospectsandbarriers,in:O.Hryniewicz,J.

Studziˇınski,A.Szediw(Eds.),EnvironmentalInformaticsandSystemsResearch, EnviroInfoWarsaw2007;The21stInternationalConferenceon“Informatics forEnvironmentalProtection”,Warsaw,Poland,September12–14,2007,vol.

1,Shaker,Aachen,2007,pp.35–42.

[11]Deutsches Institutfür Normunge.V., Environmental Management – Life cycleAssessment– PrinciplesandFramework(ISO14040:2006); German and English version EN ISO 14040:2006, Beuth, Berlin, vol. 13.020.10, 2009.

[12]DeutschesInstitutfürNormunge.V.,EnvironmentalManagement–LifeCycle Assessment–RequirementsandGuideline(ISO14044:2006);Germanand EnglishversionENISO14044:2006,Beuth,Berlin,ICS13.020.10,2006.

[13]C.Fuchs,Theimplicationsofnewinformationandcommunicationtechnologies forsustainability,EnvironmentDevelopmentandSustainability10(3)(2008) 291–309.

[14]L.M.Hilty,A.Köhler,F.VonSchéele,R.Zah,T.Ruddy,Reboundeffectsofprogress ininformationtechnology,Poiesis&Praxis:InternationalJournalofTechnol- ogyAssessmentandEthicsofScience4(1)(2006)19–38.

[15] S. Behrendt,W. Kahlenborn, M.Feil,C. Dereje, B. Raimund,D. Ruth, S.

Michael,RareMetals,MeasuresandConceptsfortheSolutionoftheProb- lemofConflict-aggravatingRawMaterialExtraction–TheExampleofColtan, Umweltbundesamt,2007.

[16] S.Bormann,L.Plank,Underpressure,in:WorkingConditionsandEconomic DevelopmentinICTProductioninCentralandEasternEurope,WEED,Berlin, 2010.

[17]J.Chan,C.Ho,Thedarksideofcyberspace,in:InsidetheSweatshopsofChina’s ComputerHardwareProduction,WEED,Berlin,2008.

[18]S.Prakash,A.Manhart,Socio-economicassessmentandfeasibilitystudyon sustainablee-wastemanagement inGhana,CommissionedbytheInspec- torateof theMinistryofHousing,Spatial PlanningandtheEnvironment of the Netherlands (VROM-Inspectorate) and the Dutch Association for the Disposal of Metal and Electrical Products (NVMP), Freiburg, 2010, http://www.oeko.de/oekodoc/1057/2010-105-en.pdf(accessed2011-03-03).

[19]A. Magashi, M. Schluep, e-Waste Assessment Tanzania, UNIDO e-waste initiative for Tanzania, Final Report, 2011, http://ewasteguide.info/files/

Magashi 2011CPCT-Empa.pdf(accessed2011-03-04).

[20]K.Brigden, I. Labunska, D. Santillo,P. Johnston,Chemical contamination at e-waste recycling and disposal sites in Accra and Korforidua,Ghana, Amsterdam, 2008, http://www.greenpeace.org/raw/content/international/

press/reports/chemical-contamination-at-e-wa.pdf(accessed2011-03-04).

[21]R.Widmer,H.Oswald-Krapf,D.Sinha-Khetriwal,M.Schnellmann,H.Böni, Globalperspectivesone-waste,EnvironmentalandSocialImpactsofElectronic WasteRecycling,EnvironmentalImpactAssessmentReview25(5)(2005) 436–458.

[22]R.Kahhat,E.Williams,Productorwaste?Importationandend-of-lifeprocess- ingofcomputersinPeru,EnvironmentalScience&Technology43(15)(2009) 6010–6016.

[23]The Climate Group, GeSI, Global e-Sustainability Initiative, SMART 2020: Enablingthe LowCarbon Economy inthe InformationAge,2008, http://www.smart2020.org/assets/files/02Smart2020Report.pdf (accessed 2011-03-04).

[24]D.Mocigemba,Sustainablecomputing,Poiesis&Praxis:InternationalJournal ofTechnologyAssessmentandEthicsofScience4(3)(2006)163–184.

[25]S.Abenius,GreenIT&GreenSoftware–timeandenergysavingsusingexisting tools,in:V.Wohlgemuth,B.Page,K.Voigt(Eds.),EnvironmentalInformatics andIndustrialEnvironmentalProtection:Concepts,MethodsandTools,Envi- roInfo2009;23rdInternationalConferenceonInformaticsforEnvironmental Protection;Proceedingsofthe23rdInternationalConferenceEnvironmental Informatics–InformaticsforEnvironmentalProtection,SustainableDevelop- mentandRiskManagement,September09–11,2009,HTWBerlin,University ofAppliedSciences,Germany,Vol.1:SessionsfürBand1,Shaker,Aachen,2009, pp.57–66.

[26]F.Albertao,J.Xiao,C.Tian,Y.Lu,K.Q.Zhang,C.Liu,Measuringthesustainabil- ityperformanceofsoftwareprojects,in:IEEEComputerSociety(Ed.),2010 IEEE7thInternationalConferenceone-BusinessEngineering(ICEBE2010), Shanghai,China,2010,pp.369–373.

[27]F. Albertao, Sustainable Software Engineering, 2004, http://www.scribd.

com/doc/5507536/Sustainable-Software-Engineering#about(accessed2010- 11-30).

[28]H.-K.Arndt,S.Lau,A.Strehl,Sustainabilityofinformationandcommunica- tionsystems(ICS),in:V.Wohlgemuth,B.Page,K.Voigt(Eds.),Environmental InformaticsandIndustrialEnvironmentalProtection:Concepts,Methodsand Tools,EnviroInfo2009; 23rdInternationalConferenceonInformatics for