traditionaltokyohotn0744

TheFourLevelsofRequirementsEngineeringforandinDynamicAdaptiveSystemsDanielM.
Berry1dberry@uwaterloo.
ca,BettyH.
C.
Cheng2chengb@cse.
msu.
edu,andJiZhang2zhangji9@cse.
msu.
edu1SchoolofComputerScienceUniversityofWaterlooWaterloo,ON,Canada2DepartmentofComputerScienceandEngineeringMichiganStateUniversityEastLansing,MI,U.
S.
A.
Abstract.
Thispaperarguesthattherearefourlevelsofrequirementsengineer-ingforandinadynamicadaptivesystem:(1)byhumans,forthegeneralbehaviorofthesystem,(2)bythesystemitself,wheneveritisadaptingbasedonchangestoitsenvironment,(3)byhumans,todecidewhen,how,andwherethesystemistoadapt,and(4)byhumans,doingresearchaboutadaptivesystems.
1OverviewRecently,asignicantamountofefforthasbeendevotedtodevelopingtechnologiestosupportdynamicadaptivesystems(DASs)[e.
g.
,1–3].
ADASisacomputer-basedsys-tem(CBS)thatiscapableofrecognizingthattheenvironmentwithwhichitsharesaninterfacehaschangedandthatiscapableofchangingitsbehaviortoadapttothechang-ingconditions.
MuchoftheinterestinDASsismotivatedbytheincreasingdemandforpervasive,mobile,andautonomiccomputing.
1.
1MotivationWehadnoticedthatREisalwaysaboutinputandresponses.
Thatis,REdetermines(1)thekindsofinputasystemmaybepresentedand(2)thesystem'sresponsestotheseinputs.
ADAS,S,isdoingREatruntime.
Thatis,Sisdetermining,asitisexecuting,(1)thekindsofinputSmaybepresentedand(2)S'sresponsestotheseinputs.
However,thisrun-timeREisnottheonlyREdoneaboutS.
HumansaredoinglotsofREaboutSandaboutS'sownRE!
Basedonthisthinking,wecharacterizethevariousREsthataretakingplaceforandinDASs.
1.
2Adapt-ReadySystemsLetSAR("AR"for"adapt-ready"[4])beaDASoperatingondomainD(i.
e.
,itsinputspace).
AtargetprogramSiofSARisaprogramexhibitingoneofthebehaviorsthatSARcanadoptafteradapting.
Si'sdomainisDi,andthesetofalltargetprogramssupportedbySARisS.
LettheinitialtargetprogramofSARbecalledS0.
Eachindexishouldberegardedasanameforsometargetprogram.
Theonlysemanticsthatcanbederivedfromthenumericalorderoftheindicesisthetimehistoryoftargetprograms.
1.
3FourLevelsofREThisnotearguesthattherearefourlevelsofrequirementsengineering(RE)comingintoplayforandinSAR.
Theyarelistedinorderofincreasingmetaness;thatis,Levelj+1REmakesdecisionsaboutthesubjectmatterofLeveljRE.
Thelevelindicesdonotindicatetheorderofoccurrence.
Ofcourse,otherdecompositionsintolevelsarepossible.
1.
Level1REisthatdonebyhumansforallthetargetprogramsinS,todetermineDiforeachSi∈SandSi'sreactiontoeachinputinDi.
Systeminvariants,whichaf-fecttheotherlevels,shouldbeidentiedatthislevel.
(Spaceconsiderationspreventelaborationonthissubject[5].
)2.
Level2REisthatdonebySARduringitsownexecutioninordertodeterminefromthelatestinputthatitmustadaptandtodeterminewhichSi∈Stoadopt.
3.
Level3REisthatdonebyhumanstodetermineSAR'sadaptationelements,whichallowSARtodotheadaptationembodiedintheLevel2RE.
4.
Level4REisthatdonebyhumanstodiscoveradaptationmechanismsingeneral.
Adaptationelementsincludedetectionandmonitoringtechniques,decision-makingprocedures,andadaptivemechanisms.
ForagivenSAR,itispossiblethattheLevels1,3,and4REs,donebyhumans,bedoneconcurrently;thatis,thehumanrequirementsengineersforSARwillneedtodeterminethesetoftargetprograms,themethodforchoosingamongthem,andgeneralmonitoringandadaptationtechniquessimultaneouslyinordertoproduceacoherentsystem.
Itispossiblealsothatthesehuman-appliedRElevelsberevisitedduringSAR'slife.
Thatis,SARmaybepresentedtotallyunanticipatedinputI∈D,suchthatSAR'sLevel2REfailstoadapt.
Perhaps,SARinformstheuserthatSARcannotadapttoI.
Perhaps,theusermustnoticethatSARisnotmeetingitsrequirements.
Then,additionalLevel1REmustbedonetodetermineatleastonenewtargetprogram,SI,thathasIinitsdomainandthatrespondscorrectlytoI.
AdditionalLevel3REmustbedonetoreviseSAR'sadaptationmechanismsothatwhenSARisrunagainwithinputI,SARdoesanewLevel2REinordertoadapttotheinputI.
Perhaps,inaddition,someLevel4REshouldbedonetodeterminebetterwaystodealwithunanticipatedinput.
2ExampleForexample,inthehistoryoftheadaptive,assistivee-mailsystemdevelopedbyFickasetal[6,7]tohelpbrain-injuredpatientsimprovetheirsocialconnectedness,onecanseeexamplesofallfourlevelsofRE.
Foreachitembelow,theparenthesizedlistgivesthesectionsofreference[7]describingtheitem'sRElevel.
–Level1REistheworkdonebyFickasetaltodetermineallpossiblee-mailfeaturesanduserinterfaces(UIs)tobesupportedbyanyversionofthee-mailsystemforacognitivelydisabledperson.
(OutermostSection5andSection5.
5)–Level3REistheworkdonebyFickasetaltodeterminethecategoriesofuserstobehelpedbythesystem,howtorecognizeauser'scategorybyhisorherinput,andtheappropriatecollectionoffeaturesforeachcategoryofuser.
ThisREwasdonebyacombinationofinterviewsofpatientsandanalysisbycaretakingexpertsandcomputingexperts;patientgoalswerematchedtoskillsneedtoachievethemandthentofeaturesrequiringthoseskills.
DoingthisREledto(1)thediscoveryoftheneedfore-mailfeaturesandUIsnotanticipatedinthepreviousLevel1REeffortand(2)theinventionoftheseadditionale-mailfeaturesandUIs,i.
e.
,someadditionalLevel1RE.
(Sections5.
1,5.
4,and5.
5)–Level2REistheworkdoneduringrunsofthee-mailsystem,asitmonitorsauser'sinputanddetermineswhenitistimetochangethee-mailsystem'sUIandbehaviortoappeartotheuserasanewe-mailprogram.
Ifthee-mailsystemcannotadapttoauserorFickasetaldeterminethattheuser'se-mailingisdeterioratingorthattheuserisbehavinginunanticipatedwaysthatarenotdetectedbytherun-timemonitoring,thenFickasetalinterveneanddomoreLevel1andLevel3RE,especiallythatinvolvingpersonalinterviewsofthepatient.
(Sections5.
2and5.
3)–Level4REisalltheresearchdonebyFickasetalandothers[8,9]inrequirementssatisfactionmonitoringandadaptation,requirementsdeferment,personalandcon-textualRE,etc.
,i.
e.
,whatFickasetaldescribeandciteintheirpapers[6,7].
(Sec-tion5.
5andReferences)Notethatinthisexampleandingeneral,Level3REwillhappenbeforeLevel2REsimplybecauseitisLevel3REthatdeterminestheLevel2REthatSARdoesduringitsexecution.
WhileinanygivenSARtheboundariesbetweenLevels1,2,and3REareprecise,inahistoryofversionsofSAR,asthehumanrequirementsengineersunderstandbettertheadaptationsthatneedtobemade,workmayshiftfromLevels1and3RE,donebyhumans,toLevel2RE,donebythenextversionofSAR.
3LevelsofRE3.
1Level1Level1REresemblesthetraditionalREthatisdoneforanyCBS.
ThisREinvolves1.
elicitingandanalyzinginformationaboutthedomainDofSAR,2.
decidingthesetofallfeaturesofanytargetprogramtobeadoptedbySARandtheirfunctionalities,3.
decidingthesetofalltargetprogramstobeadoptedbySARandtheirfunctionali-ties,and4.
specifyingthefunctionalitiesofalltargetprogramspresentedbySAR.
AwidevarietyofstandardmethodsareavailableforthisRE[e.
g.
,10–12].
3.
2Level2Level2REiswhatSARdoeswhenitgetsinputnotinthedomainofitscurrenttargetprogram.
SARmustdeterminewhichtargetprograminSitshouldadoptnext.
ThatthisbehaviorisREcanbeseenifoneconsiderswhatSARisdoing.
SupposeSARcurrentlyhasadoptedthetargetprogramSi,anditscurrentinputIisnotinDi.
Then,SAReffectively1.
determinesfromIhowitsnewdomainDi+1differsfromDi,2.
determineswhichofitstargetprograms,Si+1,toadoptnext,and3.
modiesitsownbehaviortoadoptSi+1asitscurrenttargetprogram.
Ofcourse,SARmusthavesomemonitoringcodetokeeptrackofenvironmentalchangesasreectedinitsinput.
SARmusthavecodethatdetermineswhichofitstargetprogramstoadoptasafunctionofdetectedenvironmentalchanges.
Finally,SARmusthavesomewhereinitscode,foreachtargetprogramSj,eitherthecodeforSjorcodetondthecodeforSj,e.
g.
,inalibrary.
3.
3Level3Level3REisprobablythemostdifculttoachievebecauseitrequiresassessingwhatSARshoulddoatthemetalevel,thatis,howcanwemakeSARdoitsLevel2RE.
Level3REinvolvesguringouthowtogetSARto1.
determinefromIhowitsnewdomainDi+1differsfromDi,2.
determinewhichofitstargetprograms,Si+1,toadoptnext,and3.
modifyitsownbehaviortoadoptSi+1asitscurrenttargetprogram.
DoingthisRErequireshavingdeterminedprogram-testablecorrespondencestoen-vironmentalchangesthattriggeradaptation.
Therequirementsengineerswillhavetoexplorerepresentationsfor1.
thepossiblenewdomainswiththeircorrespondingenvironmentalconditions,2.
thepossibleadaptivereactionstonewinputs,and3.
thetestableconditionsunderwhicheachnewadaptivereactionistobeapplied.
By"representation",weallowanyschemefromwhichspecicadaptivereactionscanbederived,perhapsbyinstantiation,parameterapplication,mapping,reconguration[6,7],tablelookup,recompositionofnewcomponents[1],formulaorspecicationgeneration,etc.
3.
4Level4Level4REisessentiallytheresearchintoadaptationmechanisms.
Adaptationmecha-nismshavebeendevelopedfortheapplicationlevel[e.
g.
,4,13–15],middleware[e.
g.
,16–20],andoperatingsystems[e.
g.
,21,22].
4AnotherExampleMartinFeatherdescribesadegeneratecaseofanadaptivetoolthathehaswrittenforhimselfastheonlyuser.
Hehasinsertedassertstatementsintothecodeofthetool.
Eachsuchassertstatementcausesarun-timebreakwhenitslogicalexpressionevaluatestofalse.
Eachsuchassertstatementis,ineffect,arequirementspecicationdescribinganassumedpropertyofthetool'sinputorofavaluecalculatedbythetoolinresponsetosomeinput.
Often,theviolationofanassumptionpointstoarequirementschange;heisusingthetoolinawayhehadnotanticipatedandtowhichtheexistingcodeisnotpreparedtorespondinareasonableway.
Occasionally,theviolationindicatesafeatureinteractionhedidnotanticipate.
Ineithercase,Featherreactsbyanalyzingthesituationanddecidingonnewbehavior.
Heimplementsthenewbehaviorbymanuallymodifyingthecode.
Hemodiesalsotheassertstatementstoreecttheenvironmentalassumptions.
Inthiscase,nearlyalloftheREineachofthefourlevelsisdonebyFeather,theuser–implementerhimself.
TheonlyexceptionisthepartofLevel2REthatdetectsthatthecurrentinputisnotinthetool'scurrentdomainandthatitistimetochangethetool'sbehavior.
TherestofLevel2REisdoneofflinebyFeather.
TheresultisthattheLevel3REisrathertrivial,asitinvolvesonlyguringoutthelogicalexpressionsoftheassertstatementsthatmonitorrequirementschanges.
5YetAnotherExampleThethirdexampleistheultimateDAS,CommanderData,ofStarTrek:NextGener-ation[23,24].
AlthoughDataisactionalcharacter,hewasconceivedandwrittentolifebytechnicallysavvywriterswhomanagedtoinfuseenoughconsistencyinhisbe-haviorsandabilitiesthatitispossibletoseehowhisbehaviorsandabilitiescouldbeprogrammed,givensufcientlypowerfulcomputers.
Ofcourse,currenttechnologicallimitationsprecludeData'sexistenceinanybutthefardistantfuture!
3ThefourlevelsofREinandforDataare:–Level1REisthatdonebyNoonianSoong,Data'sinventorandbuilder,forthegeneralbehaviorofallofhisandroids,includingData–Level2REisthatdonebyDatawhenherecognizesasituationnotcoveredbyhiscurrentprogrammingandpastlearning:Hesimulatesatpositroniccomputer'sspeedallsortsofrandomlygeneratedscenarioscommencingwiththecurrentsitua-tion;hechoosesandrememberstheonewiththebestoutcome;callthissimulationfollowedbyrememberingadaptationandlearning–Level3REisthatdonebyNoonianSoongtodeterminehowDataadaptsandlearns–Level4REistheresearchdonebyNoonianSoongtoimproveDataandotherandroids,e.
g.
,todeviseanemotionchipThisclassicationoftheREinvolvedinDASsandthisclassicationofthelevelsofREinthectionalultimateDAShighlightthefundamentalbarrierthatmustbe3IfMoore'slawcontinuestoholdforthenext250years,Datamightjustbepossible!
conqueredbeforeDASscanbecometrulyadaptable.
Sincefortheforeseeablefuture,softwareisnotabletothinkandbetrulyintelligentandcreative,theextenttowhichaDASSARcanadaptislimitedbytheextenttowhichtheadaptationanalystcananticipatethedomainchangestobedetectedandtheadaptationstobeperformed.
Thislimitiscalledtheenvelopeofadaptability.
ThisenvelopethusdeterminesthedomainDofSARandthesetSoftargetprogramsofSAR.
Thisenvelopeofadaptabilitycannotexceedourownadaptability.
Whileweareadaptable,wedonotknowhowweareadaptable,andthuswecannotprogramsoftwaretobeevenasadaptableasweare.
Therefore,SARwillalwaysbelessadaptablethanweare.
Inotherwords,itisnotlikelythatwewillbeabletoimplementanytimesoon,theeasyadaptabilitythatweseeintheandroidDataonStarTrekNextGenerationandtheholographicdoctoronStarTrekVoyager.
Moreover,thisadaptabilitycannothappenuntilandunlesswehumansunderstandenoughaboutourownthinkingthatweknowhowwethink,create,andadapt,andcantranslatethatknowledgeintosoftwarethattrulythinks,creates,andadapts.
Ofcourse,acleartopicforLevel4RE,i.
e.
,research,ishowanautonomicoradapt-readysystemcanadapttounanticipateddomainchangesontheywithouthumanintervention[25].
6NewComputingParadigmsThreedifferentcomputingtrendstodayneedtobuildsystemsthataredynamicallyadaptive:–autonomicsystems:Asthephenomenalgrowthofcomputersandcomputer-basedsystems(CBSs)outstripsourabilitytoprogram,run,andmaintainthem,CBSswillhavetobecomemoreandmoreselfrunning,i.
e.
autonomousandabletoruncorrectlywithonlyminimalhigh-levelinstructionfromhumans[26].
–service-centricsystems[27]:Inservice-centriccomputing,everythingisconnectedand"informationtechnologywillbeprovisioned.
.
.
,delivered.
.
.
,metered.
.
.
,man-aged.
.
.
andpurchasedasaservice".
[28].
–GRIDsystems:TheideaofGRIDcomputingistoestablishaglobalgridofcom-putingpowerbywhichausercanaccesscomputingpowerindependentlyofwheretheuserandthecomputersit,notunlikethewaytheelectricalpowergridallowaconsumertoaccesselectricpowerindependentlyofwhereeachsits[29].
Eachofthesetrendsrequireshighlydynamicadaptivesystems.
7What'sNextAswemoveforwardwithdecreasingcostsforCBSs;withincreasingdemandformobile,heterogeneous,andpervasivecomputing;andwithincreasinginterestinau-tonomicsystems[e.
g.
,25,26],theneedforDASswillincrease.
Currently,muchoftheefforthasfocusedonhowtomakelegacysystemsadaptive.
Aswemovetowardsanadaptivesoftwareparadigm,suchasforthenewcomputingparadigmsofSection6,weanticipatethattheadaptabilityenvelopewillexpandsincetheREatLevel1willexpandtoincludeREatLevels3and4.
Aswemoveintothisneweraofdynamicadaptation,moreattentionisneededtoestablishthecorrectnessofsoftware,before,during,andafteradaptation.
Thusfar,wehavelargelyfocusedontheenablingtechnologiesthatprovideadaptivecapabilities.
Weneedtostepbackandensurethatassuranceissuesarebeingconsideredatall4levelsofREforDASs.
Assurancewillcontributealsotothedecision-makingprocessfordeterminingwhen,how,andwhereadaptationsshouldtakeplace.
AcknowledgmentsWethankMartinFeatherandSteveFickasforsuggestingtwoofthemainexamplesusedinthepaper.
WethankNeilMaiden,MichaelPoppleton,andtheanonymousre-viewersofpreviousversionsofthispaperforothersuggestions.
DanielBerry'sworkissponsoredinpartbyGrant(Canada)NSERC-RGPIN227055-00.
BettyCheng'sworkissponsoredinpartsbyU.
S.
O.
N.
R.
GrantN00014-01-1-0744andN.
S.
F.
GrantsCCR-9901017,EIA-0000433,andEIA-0130724.
References1.
McKinley,P.
K.
,Sadjadi,M.
,Kasten,E.
P.
,Cheng,B.
H.
C.
:Composingadaptivesoftware.
IEEEComputer(2004)56–642.
Sousa,J.
P.
,Garlan,D.
:Aura:anarchitecturalframeworkforusermobilityinubiquitouscomputingenvironments.
In:ProceedingsofthethirdWorkingIEEE/IFIPConferenceonSoftwareArchitecture.
(2002)29–433.
Adve,S.
,Harris,A.
,Hughes,C.
,Jones,D.
,Kravets,R.
,Nahrstedt,K.
,Sachs,D.
,Sasanka,R.
,Srinivasan,J.
,Yuan,W.
:Theillinoisgraceproject:Globalresourceadaptationthroughcooperation.
In:ProceedingsoftheWorkshoponSelf-Healing,Adaptive,andself-MANagedSystems(SHAMAN).
(2002)4.
Yang,Z.
,Cheng,B.
H.
,Stirewalt,R.
E.
K.
,Sowell,J.
,Sadjadi,S.
M.
,McKinley,P.
K.
:Anaspect-orientedapproachtodynamicadaptation.
In:ProceedingsoftheACMSIGSOFTWorkshopOnSelf-healingSoftware(WOSS'02).
(2002)5.
Zhang,J.
,Cheng,B.
H.
C.
:Specifyingadaptationsemantics.
In:ICSEWorkshoponSoftwareArchitecturesforDependableSystems(WADS05).
(2005)6.
Fickas,S.
:Clinicalrequirementsengineering.
In:Proceedingsofthe27thInternationalConferenceonSoftwareEngineering.
(2005)7.
Fickas,S.
,Robinson,W.
,Sohlberg,M.
:Theroleofdeferredrequirements:Alongitudinalstudy.
In:ProceedingsoftheThirteenthIEEEInternationalConferenceonRequirementsEngineering.
(2005)8.
Fickas,S.
,Feather,M.
:Requirementsmonitoringindynamicenvironments.
In:Proc.
ofthe2ndInternationalSymposiumonRequirementsEngineering.
(1995)140–1479.
Spanoudakis,G.
,Mahbub,K.
:Requirementsmonitoringforservice-basedsystems:Towardsaframeworkbasedoneventcalculus.
In:19thIEEEInternationalConferenceonAutomatedSoftwareEngineering(ASE'04).
(2004)379–38410.
Gause,D.
,Weinberg,G.
:ExploringRequirements:QualityBeforeDesign.
DorsetHouse,NewYork,NY,USA(1989)11.
Robertson,S.
,Robertson,J.
:MasteringtheRequirementsProcess.
Addison-Wesley,Harlow,England(1999)12.
Larman,C.
:ApplyingUMLandPatterns.
Secondedn.
PrenticeHallPTR,UpperSaddleRiver,NJ,U.
S.
A.
(2002)13.
David,P.
C.
,Ledoux,T.
,Bouraqadi-Saadani,N.
M.
N.
:Two-stepweavingwithreectionusingAspectJ.
In:OOPSLA2001WorkshoponAdvancedSeparationofConcernsinObject-OrientedSystems,Tampa(2001)14.
Sadjadi,S.
M.
,McKinley,P.
K.
,Stirewalt,R.
E.
K.
,Cheng,B.
H.
:Generationofself-optimizingwirelessnetworkapplications.
In:ProceedingsoftheInternationalConferenceonAutonomicComputing(ICAC-04),NewYork,NY(2004)310–31115.
Wohlstadter,E.
,Jackson,S.
,Devanbu,P.
:DADO:enhancingmiddlewaretosupportcross-cuttingfeaturesindistributed,heterogeneoussystems.
In:ProceedingsoftheInternationalConferenceonSoftwareEngineering,Portland,Oregon(2003)174–18616.
Redmond,B.
,Cahill,V.
:Supportingunanticipateddynamicadaptationofapplicationbe-haviour.
In:Proceedingsofthe16thEuropeanConferenceonObject-OrientedProgramming.
(2002)17.
Kon,F.
,Roman,M.
,Liu,P.
,Mao,J.
,Yamane,T.
,Magalhaes,L.
C.
,Campbell,R.
H.
:Monitor-ing,security,anddynamiccongurationwiththedynamicTAOreectiveORB.
In:Proceed-ingsoftheIFIP/ACMInternationalConferenceonDistributedSystemsPlatformsandOpenDistributedProcessing(Middleware'2000).
Number1795inLNCS,NewYork,Springer-Verlag(2000)121–14318.
Blair,G.
,Coulson,G.
,Blair,L.
,andP.
Grace,H.
D.
L.
,Moreira,R.
,Parlavantzas,N.
:Reec-tion,self-awarenessandself-healinginOpenORB.
In:WOSS02,Charleston,SC(2002)19.
Zinky,J.
A.
,Bakken,D.
E.
,Schantz,R.
E.
:ArchitecturalsupportforqualityofserviceforCORBAobjects.
TheoryandPracticeofObjectSystems3(1997)20.
Sadjadi,S.
M.
,McKinley,P.
K.
:ACT:AnadaptiveCORBAtemplatetosupportunantici-patedadaptation.
In:Proceedingsofthe24thIEEEInternationalConferenceonDistributedComputingSystems(ICDCS'04),Tokyo,Japan(2004)21.
Kon,F.
,Campbell,R.
H.
,Ballesteros,F.
J.
,Mickunas,M.
D.
,Nahrstedt,K.
:2K:Adistributedoperatingsystemfordynamicheterogeneousenvironments.
In:Proceedingsofthe9thIEEEInternationalSymposiumonHighPerformanceDistributedComputing,Pittsburgh,PA,U.
S.
A.
(2000)22.
Appavoo,J.
,Hui,K.
,Soules,C.
A.
N.
,Wisniewski,R.
W.
,Silva,D.
M.
D.
,Krieger,O.
,Aus-lander,D.
J.
E.
M.
A.
,Gamsa,B.
,Ganger,G.
R.
,McKenney,P.
,Ostrowski,M.
,Rosenburg,B.
,Stumm,M.
,Xenidis,J.
:Enablingautonomicbehaviorinsystemssoftwarewithhot-swapping.
IBMSystemsJournal42(2003)23.
Hanley,R.
:IsDataHumanTheMetaphysicsofStarTrek.
BasicBooks,NewYork,NY,U.
S.
A.
(1997)24.
Gresh,L.
,Weinberg,R.
:TheComputersofStarTrek.
BasicBooks,NewYork,NY,U.
S.
A.
(1999)25.
Ganek,A.
G.
,Corbi,T.
A.
:Thedawningoftheautonomiccomputingera.
IBMSystemsJournal,SpecialIssueonAutonomicComputing42(2003)26.
Kephart,J.
O.
,Chess,D.
M.
:Thevisionofautonomiccomputing.
IEEEComputer36(2003)41–5027.
Jones,S.
V.
,Maiden,N.
A.
M.
,Zachos,K.
,Zhu,X.
:Howservicecentricsystemschangetherequirementsprocess.
In:Proceedingsof11thInternationalWorkshoponRequirementsEngineering:FoundationforSoftwareQuality.
(2005)28.
Fiorina,C.
:Thebusinessofchange.
In:TIECON,Keynote.
(2001)AvailableatURL:http://www.
hp.
com/hpinfo/execteam/speeches/orina/tiecon01.
html.
29.
Buyya,R.
:Gridcomputinginfocentre(GRIDinfoware)(2005)AvailableatURL:http://www.
gridcomputing.
com/.