Sajib Mistry Athman Bouguettaya Hai Dong Economic Models for Managing Cloud Services Economic Models for Managing Cloud Services Sajib Mistry • Athman Bouguettaya (cid:129) Hai Dong Economic Models for Managing Cloud Services 123 SajibMistry AthmanBouguettaya SchoolofInformationTechnologies SchoolofInformationTechnologies UniversityofSydney UniversityofSydney Sydney,NSW,Australia Sydney,NSW,Australia HaiDong SchoolofScience RMITUniversity Melbourne,VIC,Australia ISBN978-3-319-73875-8 ISBN978-3-319-73876-5 (eBook) https://doi.org/10.1007/978-3-319-73876-5 LibraryofCongressControlNumber:2017964374 ©SpringerInternationalPublishingAG2018 Thisworkissubjecttocopyright.AllrightsarereservedbythePublisher,whetherthewholeorpartof thematerialisconcerned,specificallytherightsoftranslation,reprinting,reuseofillustrations,recitation, broadcasting,reproductiononmicrofilmsorinanyotherphysicalway,andtransmissionorinformation storageandretrieval,electronicadaptation,computersoftware,orbysimilarordissimilarmethodology nowknownorhereafterdeveloped. Theuseofgeneraldescriptivenames,registerednames,trademarks,servicemarks,etc.inthispublication doesnotimply,evenintheabsenceofaspecificstatement,thatsuchnamesareexemptfromtherelevant protectivelawsandregulationsandthereforefreeforgeneraluse. Thepublisher,theauthorsandtheeditorsaresafetoassumethattheadviceandinformationinthisbook arebelievedtobetrueandaccurateatthedateofpublication.Neitherthepublishernortheauthorsor theeditorsgiveawarranty,expressorimplied,withrespecttothematerialcontainedhereinorforany errorsoromissionsthatmayhavebeenmade.Thepublisherremainsneutralwithregardtojurisdictional claimsinpublishedmapsandinstitutionalaffiliations. Printedonacid-freepaper ThisSpringerimprintispublishedbytheregisteredcompanySpringerInternationalPublishingAGpart ofSpringerNature Theregisteredcompanyaddressis:Gewerbestrasse11,6330Cham,Switzerland Tomyparents,Sunil andMili,andmywife, Pragga,andmydaughter,Aapti. SajibMistry Tomywifeandbestcompanion andfriend, Malika. AthmanBouguettaya Tomyparents,JianwenDongandAiying Yang. HaiDong Foreword Cloud computing is inexorably making large inroads to becoming the technology of choice among small and large businesses via provisioning IT infrastructure resources and hosting applications. A significant advantage of cloud computing is itseconomicbenefitstobothserviceconsumersandproviders.Fromaconsumer’s perspective, cloud computing is a model for providing computing resources as a location-independent and highly scalable service that is acquired on demand with little or no fixed capital investment. Cloud is ideal for those organizations with fluctuatingcomputationalresourcedemands.Itisabletofulfillthosedemandswith dynamicallyprovisionedresources.Fromaprovider’sperspective,cloudcomputing provides economies of scale through a distribution of costs among a large pool of consumers, centralization of infrastructures in locations with lower costs, and improvedresourceutilization. Cloud service management is a critical aspect of cloud computing. In cloud computing,acomplexbusinessprocesscanberealizedbyoutsourcingitsinvolved tasks to the cloud and then composing the resulting component cloud services against each of these outsourced tasks. The cloud service management from the consumers’ perspective aims to select the optimal provider for each component cloudservicetominimizeconsumers’totalcostwhilemeetingconsumers’Quality ofService(QoS)requirements.Similarly,cloudprovidersrequireaneffectivecloud service management process to achieve their business goals, e.g., specific revenue orprofitexpectations.Inthisregard,thecloudservicemanagementprocesscanbe realized by selecting and composing more profitable services to fulfill consumers’ demandswhileensuringdeliveryofthepromisedQoS.Fromhere,wecanseethat cloud computing management is mainly driven by economic factors. Employing economicmodelstodesignthecloudservicemanagementframeworkisarightway, asitperfectlyfitsthewayinwhichbusinessisconducted. There are a variety of books on the market that cover many interesting issues relatedtocloudcomputing.However,nonebutthisbookprovideacomprehensive analysisofeconomicmodelsinthelong-termcloudservicemanagementfromthe providers’perspective.Thisbookisafirstattempttodesignquantitativeandqualita- tiveeconomicmodelsforanInfrastructureasaService(IaaS)provider.Itprovides vii viii Foreword an efficient, market-driven, and competitive environment for cloud providers and consumers to maximize their profit and minimize their costs, respectively, while meetingtheirlong-termQoSrequests.Thisiscrucialifwewanttounlockthefull potential of cloud computing which has largely remained closed and proprietary. This book also covers fundamental technical details concerning optimization, prediction, and machine learning models in the cloud service management. The state-of-the-arttechnologiesdescribedandthereferencesincludedinthisbookwill alsohelptheinterestedreadersgainknowledgeonthesetopics. This book can be utilized as a useful reference to anyone who is interested in theory,practice,andapplicationofeconomicmodelsincloudcomputing.Thisbook willbeaninvaluableguideforsmallandmediumentrepreneurswhohaveinvested orplantoinvestincloudinfrastructuresandservices.Overall,thisbookissuitable for a wide audience that includes students, researchers, and practitioners studying orworkinginservice-orientedcomputingandcloudcomputing. SchoolofComputerScienceandEngineering ProfessorBoualemBenatallah TheUniversityofNewSouthWales Sydney,NSW,Australia September18,2017 Preface Cloud computing is increasingly becoming the technology of choice as the next-generationplatformforconductingbusinesses.Cloudsolutionsarerepresented as services, i.e., a higher level abstraction of computing resources. An effective cloud service management framework has the potential of creating a sustainable cloud service market. It enables the wider adoption of the cloud at a greater scale and faster pace by considering economic perspectives of service consumers and providers. The long-term cloud service composition is an essential element to designing an effective management framework. The composition from the consumer’s perspective aims to minimize total costs while meeting Quality of Service (QoS) requirements by selecting the best set of cloud providers. The composition from the provider’s perspective aims to maximize profit for a long- termperiodbyselectinganoptimalsetofservicerequests.Developinganefficient andlong-termcloudservicecompositionframeworkfromtheprovider’sperspective isveryimportanttosustainingandgrowingdemandfromalargenumberofservice consumersinthemarket. In this book, we develop an economic model-driven long-term cloud service composition framework from the Infrastructure as a Service (IaaS) provider’s perspective.Theprovider-consumerrelationshipsinacloudmarketareusuallylong termandeconomicallydriven.Inthisregard,usingeconomicmodelstooptimizethe composition of service requests fits perfectly well with the way in which business is usually conducted. First, a new quantitative economic model is developed that maximizestheprovider’slong-termrevenueandprofitbyselectinganoptimalsetof IaaSrequestsinadynamiceconomicenvironment.Weproposeanewmultivariate Hidden Markov and Autoregressive Integrated Moving Average (HMM-ARIMA) model to predict various patterns of runtime resource utilization. A heuristic- based Integer Linear Programming (ILP) optimization approach is proposed to maximizetheruntimeresourceutilization.WedeployaDynamicBayesianNetwork (DBN)tomodelthedynamicpricingandlong-termoperationcost.AnewHybrid AdaptiveGeneticAlgorithm(HAGA)isdevelopedthatoptimizesanonlinearprofit functionperiodicallytoaddressthestochasticarrivalofrequests.Next,wedevelopa qualitativeeconomicmodelwhichisapreference-drivenapproachtoenactonhigh- ix x Preface level business strategies, such as service menu creation and pricing. We propose theTemporalConditionalPreferenceNetwork(TempCP-Net)torepresentthehigh- levelIaaSbusinessstrategies.Thetemporalqualitativepreferencesareindexedina multidimensionalk-dtreetoefficientlycomputethepreferencerankinginruntime. Athree-dimensionalQ-learningapproachisproposedtofindanoptimalqualitative composition using statistical analysis on historical request patterns. Finally, we propose a new multivariate approach to predict future Quality of Service (QoS) performancesofpeerserviceproviderstoefficientlyconfigureaTempCP-Net. We have evaluated the efficiency of the proposed framework using Google Cluster data, real-world QoS data, and synthetic data. Experimental results show thattheproposedcompositionframeworkefficientlymaximizestheprovider’slong- termeconomicgoalsinruntime.Theproposedmodelsinthisbookareexpectedto playasignificantroleincreatinganeconomicallyviableandstablecloudmarket. Sydney,NSW,Australia SajibMistry Sydney,NSW,Australia AthmanBouguettaya Melbourne,VIC,Australia HaiDong Acknowledgments Ioweahugedebtofgratitudetomywife,withoutwhoseloveandsupport,Iwould not have finished this book. I would also like to thank my parents, parents-in-law, brother,andsister-in-lawfortheirencouragement.Iwouldalsoliketothankallmy friendswhodirectlyorindirectlysupportedme. Sydney,NSW,Australia SajibMistry I would like to thank my family for their unwavering support during my work on thisbook. Sydney,NSW,Australia AthmanBouguettaya I would like to acknowledge with gratitude the love and support of my family membersandfriendsalongtheway. Melbourne,VIC,Australia HaiDong Theauthorsofthisbookwouldliketoextendtheirsinceregratitudeandappreciation totheircollaboratorsforthecontributiontothisbook.Inparticular,wewouldlike toacknowledgeDr.KaiQin,Dr.AbdelkarimErradi,andothercollaboratorsinthe SensorCloudServicesLaboratory(SCSLab)attheUniversityofSydney. xi