iot based agriculture as a cloud and big data service · iot based agriculture as a cloud and big...
Post on 27-Jun-2020
0 Views
Preview:
TRANSCRIPT
DOI: 10.4018/JOEUC.2017100101
Journal of Organizational and End User ComputingVolume 29 • Issue 4 • October-December 2017
Copyright©2017,IGIGlobal.CopyingordistributinginprintorelectronicformswithoutwrittenpermissionofIGIGlobalisprohibited.
IoT Based Agriculture as a Cloud and Big Data Service:The Beginning of Digital IndiaSukhpal Singh Gill, CLOUDS Lab, School of Computing and Information Systems, University of Melbourne, Melbourne, Australia
Inderveer Chana, Computer Science and Engineering Department, Thapar University, Patiala, India
Rajkumar Buyya, CLOUDS Lab, School of Computing and Information Systems, University of Melbourne, Melbourne, Australia
ABSTRACT
Cloudcomputinghastranspiredasanewmodelformanaginganddeliveringapplicationsasservicesefficiently.Convergenceofcloudcomputingwithtechnologiessuchaswirelesssensornetworking,InternetofThings (IoT)andBigDataanalyticsoffersnewapplications’of cloud services.Thispaperproposesacloud-basedautonomicinformationsystemfordeliveringAgriculture-as-a-Service(AaaS)throughtheuseofcloudandbigdatatechnologies.TheproposedsystemgathersinformationfromvarioususersthroughpreconfigureddevicesandIoTsensorsandprocessesitincloudusingbigdataanalyticsandprovidestherequiredinformationtousersautomatically.TheperformanceoftheproposedsystemhasbeenevaluatedinCloudenvironmentandexperimentalresultsshowthattheproposedsystemoffersbetterserviceandtheQualityofService(QoS)isalsobetterintermsofQoSparameters.
KEywORDSAgriculture as a Service, Autonomic Management, Big Data, Cloud Computing, Internet of Things
1. INTRODUCTION
EmergenceofICT(InformationandCommunicationTechnologies)playsanimportantroleintheagriculture sectorbyproviding services throughcomputer-basedagriculture systems (SinghandChana,2015).Buttheseagriculturesystemsarenotabletofulfilltheneedsoftoday’sgenerationduetoprocessingoflargeamountofdata,lackofimportantrequirementslikeprocessingspeed,datastoragespace,reliability,availability,scalabilityetc.andevenresourcesusedincomputer-basedagriculturesystemsarenotutilizedefficiently.Agriculture-as-a-Service(AaaS)applicationsexhibitBigdatacharacteristics.Forexample,thevolumeofagriculturedatasetcapturedbyenvironmentssuchasOpenGovernmentDataPlatformIndia(data.gov.in,2015),IndiaAgricultureandClimateDataSet(Sanghietal.),andregionallandandclimatemodellinginChina(Shangguanetal.,2012)canbeinorderof1000000recordswithsizeof3.5GB.Thedataiscominginlargedatavarietyandvolumefrombothusersintheformofimageslikedamagedcropimagesduetoweather,insectsetc.anddevicesthroughInternetofThings(IoT)sensorsandsatellites(GPSsystems)thatsendweatherrelatedimages.Asaresultofregularcapturingandcollectionofdatasets,theygrowwiththevelocity
1
Journal of Organizational and End User ComputingVolume 29 • Issue 4 • October-December 2017
2
of80.72KB/minuteormore(data.gov.in,2015).Tosolvetheproblemofexistingagriculturesystems,thereisaneedtodevelopacloud-basedservicethatcaneasilymanagedifferenttypesofagriculturerelated-databasedondifferentdomains(crop,weather,soil,pest,fertilizer,productivity,irrigation,cattle,andequipment)throughthesesteps:i)gatherdatafromvarioussensorsthroughpreconfigureddevices,ii)classifythegathereddata(heterogeneous,highvolumeofbigdata)intovariousclassesthrough analysis, iii) store the classified information in cloud repository for future use, and iv)automaticdiagnosisoftheagriculturestatus.Aslargenumberofusersareusingagriculturesystemsoperatingonlargedatasetssimultaneously,thereisaneedofhighlyscalableandelasticdistributedcomputingenvironmentsuchascloudcomputing.Inaddition,cloud-basedautonomicinformationsystemshouldbeable to identify theQoS(QualityofService)requirementsofuserrequestandresourcesshouldbeallocatedefficientlytoexecutetheuserrequestbasedontheserequirements.
Themainaimofthispaperis todesignarchitectureofAgriculture-as-a-Service(AaaS)thatmanagesvarioustypesofagriculture-relateddatabasedondifferentdomains.Thisisrealizedthroughthefollowingobjectives:i)proposeanautonomicresourcemanagementtechniquewhichisusedtoa)gathertheinformationfromvarioususersthroughpreconfigureddevices,IoTsensors,GPS(GlobalPositioningSystem),etc.b)extract theattributes,c)analyze the informationbycreatingvariousclassesbasedontheinformationreceived,d)storetheclassifiedinformationincloudrepositoryforfutureuseande)diagnosetheagriculturestatusautomaticallyandii)performresourceallocationautomaticallyatinfrastructurelevelafteridentificationofQoSrequirementsofuserrequest.
The rest of the paper is organized as follows. Section 2 presents related work of existingagriculturessystems.ProposedarchitectureispresentedinSection3.Section4presentsAutonomicResourceManagement.Sections5describetheexperimentalsetupandpresenttheresultsofevaluation.Section6presentsconclusionsandfuturescope.
2. RELATED wORK
Existingresearchreportedthatfewagriculturesystemshavebeendevelopedwithlimitedfunctionality.Relatedworkofexistingagriculturesystemshasbeenpresentedinthissection.
2.1. Existing Agriculture SystemsRanyaetal.(2013)presentedALSE(AgricultureLandSuitabilityEvaluator)tostudyvarioustypesof land to find theappropriate land fordifferent typesof cropsbyanalyzinggeo-environmentalfactors. ALSE used GIS (Global Information System) capabilities to evaluate land using localenvironmentconditionsthroughdigitalmapandbasedonthisinformationdecisionscanbemade.Raimoetal.(2010)proposedFMIS(FarmManagementInformationSystem)usedtofindtheprecisionagriculturerequirementsforinformationsystemsthroughweb-basedapproach.AuthoridentifiedthemanagementofGISdataisakeyrequirementofprecisionagriculture.Sorensenetal.(2010)studiedtheFMIStoanalyzedynamicneedsoffarmerstoimprovedecisionprocessesandtheircorrespondingfunctionalities.Furthertheyreportedthatidentificationofprocessusedforinitialanalysisofuserneeds ismandatory foractualdesignofFMIS.Zhao (2002)presentedananalysisofweb-basedagriculturalinformationsystemsandidentifiedvariouschallengesandissuesstillpendinginthesesystems.Duetolackofautomationinexistingagriculturesystem,thesystemistakinglongertimeandisdifficulttohandledynamicneedsofuserwhichleadstocustomerdissatisfaction.Sorensenetal.(2011)identifiedvariousfunctionalrequirementsofFMISandinformationmodelispresentedbasedontheserequirementstorefinedecisionprocesses.TheyidentifiedthatcomplexityofFMISisincreasingwithincreaseinfunctionalrequirementsandfoundthatthereisaneedofautonomicsystemtoreducecomplexity.Yuegaoetal.(2004)proposedWASS(Web-basedAgriculturalSupportSystem)andidentifiedfunctionalities(information,collaborativeworkanddecisionsupport)andcharacteristicsofWASS.Basedoncharacteristics,authorsdividedWASSintothreesubsystems:production,research-educationandmanagement.
Journal of Organizational and End User ComputingVolume 29 • Issue 4 • October-December 2017
3
Reddyatel.(1995)proposedGISbasedDSS(DecisionSupportSystem)frameworkinwhichSpatialDDShasbeendesignedforwatershedmanagementandmanagementofcropproductivityatregionalandfarmlevel.GISisusedtogatherandanalyzethegraphicalimagesformakingnewrulesanddecisionsforeffectivemanagementofdata.Shitalaetal.(2013)presentedmobilecomputingbasedframeworkforagriculturistscalledAgroMobileforcultivationandmarketingandanalysisofcropimages.Further,AgroMobileisusedtodetectthediseasethroughimageprocessingandalsodiscussedhowdynamicneedsofuseraffectstheperformanceofsystem.Seokkyunetal.(2013)proposedcloudbasedDiseaseForecastingandLivestockMonitoringSystem(DFLMS)inwhichsensornetworkshasbeenusedtogatherinformationandmanagesvirtually.DFLMSprovidesaneffectiveinterfaceforuserbutduetotemporarystoragemechanismused,itisunabletostoreandretrievedataindatabasesforfutureuse.TheproposedQoS-awareCloudBasedAutonomicInformationSystem(AaaS)hasbeencomparedwithexistingagriculturesystemsasdescribedinTable1.
AlltheaboveresearchworkshavefocusedondifferentdomainsofagriculturewithdifferentQoSparameters.Noneoftheexistingagriculturesystemsconsidersself-managementofresources.Duetolackofautomationofresourcemanagement,servicesbecomeinefficientwhichfurtherleadsto customer dissatisfaction. The proposed system is a novel QoS-aware cloud based autonomicinformationsystemandconsidersvariousdomainsofagricultureand,allocatesandmanagestheresourcesautomaticallywhichisnotconsideredinotherexistingagriculturesystems.
3. AGRICULTURE-AS-A-SERVICE ARCHITECTURE
Theexistingagriculturesystemsarenotabletofulfilltheneedsoftoday’sgenerationduetolackinginimportantrequirementslikeprocessingspeed,datastoragespace,reliability,availability,scalabilityetc.Evenresourcesusedincomputerbasedagriculturesystemsarenotutilizedefficiently.Tosolvetheproblemofexistingagriculturesystems, there isaneed todevelopacloud-basedautonomicinformation system that delivers Agriculture-as-a-Service. This section presents architecture ofcloud-basedautonomicinformationsystemforagricultureservicecalledAaaSthatmanagesvarious
Table 1. Comparisons of existing agriculture systems with proposed system (AaaS)
Agriculture System Mechanism QoS-aware
(Parameter) Domains Data Classification
Resource Management
Big Data
ALSE(Elsheikhetal.,2013)
Non-Autonomic Yes(Suitability) Soil Yes No No
FMIS(Nikkilaetal.,2010)
Non-Autonomic No PestandCrop No No No
WASS(Huetal.,2004)
Non-Autonomic No Productivity No No No
AgroMobile(Prasadetal.,2013)
Non-Autonomic
Yes(Dataaccuracy) Crop Yes No No
DFLMS(Jeongetal.,2013)
Non-Autonomic No Crop No Yes No
ProposedSystem(AaaS) Autonomic
Yes(Cost,Time,ResourceUtilization,Latency,ThroughputandAttackDetectionRate)
Crop,Weather,Soil,Pest,FertilizerandIrrigation
Yes Yes Yes
Journal of Organizational and End User ComputingVolume 29 • Issue 4 • October-December 2017
4
typesofagriculture-relateddatabasedondifferentdomains.ArchitectureofAaaSisshowninFigure1.QoSparameters(executiontimeandcost)mustbeidentifiedbeforetheallocationofresources.AaaSisthekeymechanismthatensuresthattheresourcemanagercanservelargeamountofrequestswithoutviolatingSLAtermsanddynamicallymanagestheresourcesbasedonQoSrequirementsidentifiedbyQoSmanager.TheservicesofAaaShasbeendividedintothreetypes:SaaS(SoftwareasaService),PaaS(PlatformasaService)andIaaS(InfrastructureasaService).InSaaS,auserinterfaceisdesignedinwhichuserscaninteractwithsystem.Anekaisa.NET-basedapplicationdevelopmentPaaS,whichisusedasascalablecloudmiddlewaretomakeinteractionbetweencloudsubsystem and user subsystem. In IaaS, an autonomic resource manager manages the resourceautomaticallybasedontheidentifiedQoSrequirementsofaparticularrequest.ThearchitectureofAaaScomprisesoftwosubsystems:i)userandii)cloud.
3.1. User SubsystemThissubsystemprovidesauserinterface,inwhichdifferenttypeofusersinteractwithAaaStoprovideandgetusefulinformationaboutagriculturebasedondifferentdomains.Ninetypesofinformationofdifferentdomainsinagriculturehasbeenconsidered:crop,weather,soil,pest,fertilizer,productivity,irrigation, cattle, andequipment.Usersarebasicallyclassified in threecategories: i) agricultureexpert,ii)agricultureofficer,andiii)farmer.TheagricultureexpertsharesprofessionalknowledgebyansweringfarmerqueriesandupdatestheAaaSdatabasebasedonthelatestresearchdoneinthefieldofagriculturewithrespecttotheirdomain.Agricultureofficersarethegovernmentofficialsthatprovidethelatestinformationaboutnewagriculturepolicies,schemes,andrulespassedbythegovernment.FarmerisanimportantentityofAaaSwhocantakemaximumadvantagebyaskinghisqueriesandgettingautomaticreplyafteranalysis.Userscanmonitoranydatarelatedtotheirdomainandgettheirresponsewithoutvisitingtheagriculturehelpcenter.ItintegratesthedifferentdomainsofagriculturewithAaaS.Thequeriesreceivedfromuser(s)areforwardedtocloudrepositoryforupdatesandresponsesendsbacktoparticularuserontheirpreconfigureddevices(tablets,mobilephones,laptopsetc.)viainternet.
3.2. Cloud SubsystemThis subsystemcontains theplatform inwhich agriculture service is hostedon a cloud.Detailsabout users and agriculture information are stored in a cloud repository in different classes fordifferentdomainswithuniqueidentificationnumber.Theinformationismonitored,analyzed,andprocessedcontinuouslybyAaaS.Theanalysisprocessconsistsofvarioussubprocesses:selection,datapreprocessing,transformation,classificationandinterpretationasshowninFigure1.Differentclassesforeverydomainandsubclassesforfurthercategorizationofinformationhavebeendesigned.Instoragerepository,userdataiscategorizedbasedondifferentpredefinedclassesofeverydomain.Thisinformationisfurtherforwardedtoagricultureexpertsandagricultureofficersforfinalvalidationthroughpreconfigureddevices.Further,anumberofuserscanusecloud-basedagricultureservicesotheQoSmanagerandautonomicresourcemanagerincloudsubsystemhavebeenintegrated.QoSmanageridentifiestheQoSrequirementsbasedonthenumberandtypeofuserqueriesasdiscussedinpreviousresearchwork(Jeongetal.,2013;SinghandChana,2015;Singhetal.,2015).BasedonQoSrequirements,autonomicresourcemanageridentifiesresourcerequirementsautomaticallyandallocatesandexecutestheresourcesatinfrastructurelevel.Performancemonitorisusedtoverifytheperformanceofsystemandalsomaintainitautomatically.Ifthesystemwillnotbeabletohandletherequestautomaticallythenthesystemgeneratesanalert.
3.2.1. Cloud-Based Agriculture ServiceCloud-basedagricultureserviceprovidesauserplatformthroughwhichusercanaccessagricultureserviceasshowninFigure2.Firstly,agricultureserviceallowsusertocreateprofileforinteractionwithAaaS.Afterprofilecreation,theuserisrequiredtoprovidehispersonaldetailsalongwiththe
Journal of Organizational and End User ComputingVolume 29 • Issue 4 • October-December 2017
5
detailsofinformationdomain.AaaSanalysestheinformationtoverifywhetherthedataiscompleteornotforfurtherprocessingbyperformingvariouschecks.Furtherdataisprocessedandredundancyofdataisremovedanddataisusedtoselectdomaintowhichdatabelongs.Informationisclassifiedproperlyinorderwithuniqueidentificationnumber.Thisinformationisforwardedtoagricultureexpertsandagricultureofficersforfinalvalidationthroughpreconfigureddevices.Aftersuccessfulvalidationofinformation,itisstoredinAaaSdatabase.Ifuserwantstoknowtheresponseoftheirquery,thensystemwillautomaticallydiagnosetheuserqueryandsendtheresponsebacktothatuser.
Figure 1. Agriculture-as-a-Service architecture
Figure 2. Functional aspects of AaaS
Journal of Organizational and End User ComputingVolume 29 • Issue 4 • October-December 2017
6
3.2.2. Detailed MethodologyAaaSallowsuserstouploadthedatarelatedtodifferentdomainsofagriculturethroughpreconfigureddevicesandclassifiedthembasedonthedomainsspecifiedindatabase.SubtasksofinformationgatheringandprovidedinAaaSare:i)selection,ii)preprocessing,iii)transformation,iv)classificationandv)interpretation.Inselection,targetdatasetsarecreatedbasedontherelevantinformationthatwillfurtherbeconsideredforanalysisinnextsubprocess.Inpreprocessing,differentusershavedifferentinformationregardingagriculture.Todevelopafinaltrainingset,thereisneedofpreprocessingstepsbecausedatamightcontainsomemissingsampleornoisecomponents.InAaaS,datapreprocessingcontainsfourdifferentsubprocesses:i)datacleaning,ii)dataintegration,iii)dataconversionandiv) data reduction. Data transformation provides an interface between data analysis subprocess(classification)anddatapreprocessing.Afterdatapreprocessing,thisprocessconvertsthelabeleddataintoadequateformatsuitableforclassification.Inclassification,AaaSclassifytheagricultureinformationofdifferentusersofdifferentdomainsbasedontheextracteddata.K-NN(k-NearestNeighbor)classificationmechanismhasbeenused in this researchwork to identify thedifferentclasslabelsofusers.K-NNissupervisedmachinelearningtechniquewhichisusedtoclassifytheunknowndatausingtrainingdatasetgeneratedbyit.K-NNusedtoidentifytheproductivitylevelthroughTrainingInstanceDataset(TID).Figure3describestheK-NNAlgorithm.
InK-NNalgorithm,distanceiscomputedfromonespecificinstancetoeverytraininginstancetoclassifythatunknowninstance.Bothk-nearestneighborandkminimumdistanceisdeterminedandoutputclasslabelisidentifiedamongkclasses.Duringtrainingphase,K-NNAlgorithmutilizestrainingdata.Figure4illustratestheclassificationprocessusedinthisresearchwork.
K-NNmodelisusedtoidentifytheproductivitylevelthroughTrainingInstanceDataset(TID).Fivelevelsofproductivity(A-E)havebeenfixedasshowninTable2.Thelevel‘A’indicatestheproductivityisveryhighwhilelevel‘E’indicatestheproductivityisverylow.Basedonthegiveninformation,TIDidentifiestheclassinwhichgivendatabelongs.
TestdataisaninputofthismodelanditiscomparedwithTIDandidentifiestheclassinwhichdatalaidusingfollowingrule:
Rule:If {CropName˄ Temperature˄ SoilTexture˄ Season˄ Pesticide˄ Fertilizer}thenProductivity
Thefinalstepistointerprettheagriculturedatasubmittedbydifferentusersofdifferentdomainswhichhelpsusertounderstandtheclassifieddatasets.AaaSiscapabletodiagnosetheagriculturestatusbasedontheinformationenteredbyuserandsendthediagnosedagriculturestatustoparticularuser
Figure 3. Pseudo code of K-NN algorithm
Journal of Organizational and End User ComputingVolume 29 • Issue 4 • October-December 2017
7
automatically.Sixattributeshavebeenconsidered:CropName,Temperature,SoilTexture,Season,PesticideandFertilizerandoneoutput:Productivity.Basedonthesesixattributes,AaaSdesignsrules.ValuesforsixvariablesareconsideredasTID.Forexample,refertoTable3.
AaaSusestheruleshowninTable3tofindtheproductivitylevelusingTID(seeTable4).Similarly,anytypeofqueryrelatedtodifferentdomainscanbeaskedbyusersandAaaSexecutes
theuserqueryandsendresponsebacktoparticularuserautomaticallybasedontherulesdefinedinAaaSdatabase.ThroughAaaS,userscaneasilydiagnosetheagriculturestatusautomatically.
3.2.3. Infrastructure Management (IaaS)EfficientmanagementofinfrastructureincloudismandatorytomaintaintheperformanceoftheAgri-Info.Itcomprisesoftwosubunits:QoSManagerandResourceManager.
Figure 4. Classification process
Table 2. Productivity Levels
Productivity Level Description
A VeryHighProductivity
B HighProductivity
C NeutralProductivity
D LowProductivity
E VeryLowProductivity
Table 3. User wants to retrieve the productivity level using AaaS
User Query
Crop Name Temperature Soil
Texture Season Pesticide Fertilizer Productivity
Soybean 21-27°C SlityLoamClay Winter Organochlorine Urea ?
Journal of Organizational and End User ComputingVolume 29 • Issue 4 • October-December 2017
8
3.2.3.1. QoS ManagerUsersubmitsarequesttoAgri-Infotoretrievesomespecificagriculturerelatedinformation.Agri-InfoidentifiestheQoSparametersrequiredtoprocesstheuserrequestthroughanalysisbasedonuserrequest.BasedonthekeyQoSrequirementsofaparticularuserrequest,theQoS Managerputstheuserrequestintocriticalandnon-criticalqueuesthroughQoSassessment.ForQoSassessment,QoS Managerwillcalculatetheexecutiontimeofuserrequestandfindtheapproximateuserrequestcompletion time. If the completion time is lesser than the desired deadline then it will executeimmediatelywiththeavailableresourcesandreleasetheresource(s)backtoresourcemanagerforanother execution otherwise calculate extra number of resources required and provide from thereservedstockforcurrentexecution.3.2.3.2. Resource ManagerFurther,tworesourceschedulingpolicies(SinghandChana,2015)areusedtoscheduletheresourcesforexecutionofuserqueries:timebasedandcostbasedschedulingpolicy.Time based scheduling policyworksasperfollowing:First,theallocationagentbeginstocomputetheDeadlineTimeoftheuserrequestinthegivenbudget.Allocateresourcesbasedontime,theuserrequestwhichhasshortestDeadlineTimewillexecutefirst.Ifthetworequestshavesamedeadlinetimethenthatrequestwillexecutefirstthathaslesserexecutiontime.TheallocationagentthenschedulesalltherequestswithsmallestexecutiontimerequesttotheresourcesthatprovidehighQoS.TherulesfortimebasedschedulingpolicyaredescribedinTable5alongwiththeirconditions.
Cost based scheduling policyworksasperfollowing:First,theallocationagentbeginstocomputethecostofeachrequestthensort,asthepriorityisgiventotherequestwhichhasmaximumbudget.Ifthetworequestshavesamebudgetthenthatrequestwillexecutefirstthathaslesserexecutiontime.TheallocationagentthenschedulesalltherequestswithhighbudgetrequesttotheresourcesthatprovidehighQoS.Finally,allotherrequestsarescheduledontheavailableresourcesset.TherulesforcostbasedschedulingpolicyaredescribedinTable6alongwiththeirconditions.
4. AUTONOMIC RESOURCE MANAGEMENT
WorkingofautonomicelementofAgri-InfoisbasedonIBM’sautonomicmodelthatconsidersfourstepsofautonomicsystem:i)monitor,ii)analyze,iii)planandiv)executeasshowninFigure1.The
Table 4. AaaS response utilized to in order to find the productivity level using TID
AaaS Response
Crop Name Temperature Soil
Texture Season Pesticide Fertilizer Productivity
Soybean 21-27°C SlityLoamClay Winter Organochlorine Urea C
Table 5. Rules of time based resource scheduling
Request Pending Urgency Add Resource Request
Yes Yes Reserve Submit
Yes No Available Submit
No - - Finish
Journal of Organizational and End User ComputingVolume 29 • Issue 4 • October-December 2017
9
objectiveofresourceprovisioninginautonomicresourcemanagementistoprovisiontheresourcestoprocessuserrequests.Therequestssubmittedshouldbeexecutedwithintheirbudgetanddeadline.Requestssubmittedbyusertoresourceprovisionerarestoredasbulkofworkloadsfortheirexecution.AllthesubmittedworkloadsareanalyzedbasedontheirQoSrequirements.Basedonimportanceoftheattribute,weightsforeverycloudworkloadarecalculated.Afterthat,workloadsareclusteredbasedonk-meansbasedclusteringalgorithmforbetterresourceprovisioning(Singhetal.,2015).Ifthevalueofworkloadsexecuteswithindeadlineandbudgetand[ResourceConsumptionandRequestsMissedislesserthanThresholdValue]thenitwillprovisionresourcesotherwisegeneratealertforanalysestheworkloadagain.
Aftersuccessfulprovisioningofresources,ResourceScheduler(RS)takestheinformationfromtheappropriateworkloadafteranalyzingthevariousworkloaddetailswhichuserrequestdemanded(SinghandChana,2015).KnowledgeBasecontainsdetailsofalltheresourcesavailableinresourcepoolandreserveresourcepool.BasedonCloudconsumerdetails,RSassignsresourcesandexecutesCloudworkloads.Duringexecutionofaparticularcloudworkload,theResourceExecutor(RE)willcheckthecurrentworkload.Iftheresourcesaresufficientforexecutionthenitwillcontinuewithexecutionotherwiserequestformoreresources.IfthevalueofResourceConsumptionandRequestsMissedislesserthanthresholdvalue,thenREwillexecuteworkloadsotherwiseREwillgeneratealert.AftersuccessfulexecutionofCloudworkloads,REreleasesthefreeresourcestoresourcepoolandREisreadyforexecutionofnewcloudworkloads.Duringexecutionofuserrequests,performanceismonitoredcontinuouslyusingsubunitperformancemonitortomaintaintheefficiencyofAgri-Infoandgeneratesalertincaseofperformancedegradation.Alertscanbegeneratedintwoconditionsgenerally:i)ifresourceconsumptionismorethanthresholdvaluesofresourceconsumptiontoexecuteuserrequest(Action:Reallocatesresources)andii)ifthenumberofmissedrequestsaregreaterthanthethresholdvalue(Action:PredictQoSRequirementsAgain).Sameactionisperformedtwice,ifAgri-Infofailstocorrectitthensystemwillbetreatedasdown.Componentsofautonomicsystemaredescribedbelow:
4.1. SensorsSensorsgettheinformationaboutperformanceofothernodesusinginthesystemandtheircurrentstate.Firstly,theupdatedinformationfromprocessingnodesistransfertomanagernodethenmanagernode transfers this information to sensors.Updated information includes informationaboutQoSparameters(executiontime,executioncostandresourceutilizationetc.).
4.2. MonitorInitially, Monitors are used to collect the information from sensors for monitoring continuouslyperformancevariationsbycomparingexpectedandactualperformance,andmonitorsthevalueof
Table 6. Rules of cost based resource scheduling
Request Pending RA > 0 Et > Wd BA > Pr Status
Yes True True True AddResource
Yes False True True AddResource
No - - - Finish
Yes True False True Finish
Yes True True False Finish
RA
= Resource Available, Et
= Estimated Time, Pr
= Resource Price, Wd
= Desired Deadline and BA
= Available Budget. Details of both time and cost based scheduling policy is given in previous research work (Singh and Chana, 2015).
Journal of Organizational and End User ComputingVolume 29 • Issue 4 • October-December 2017
10
resourceconsumptionandmissedrequests.ActualinformationaboutperformanceisobservedbasedQoSparametersandtransfersthisinformationtonextmoduleforfurtheranalysis.
4.3. Analysis and PlanAnalyzeandplanmodulestartanalyzingtheinformationreceivedfrommonitoringmoduleandmakeaplanforadequateactionsforcorrespondingalert.FollowingformulaisusedtocalculateResourceConsumption(Equation1):
ResourceConsumption
i
n
� ==∑1
Actual�Resource�Usage
Predicted�Reesource�Usage
(1)
whereActual Resource Usage isusageofresourcetoexecuteparticularnumberofuserrequestsand Predicted Resource Usage isresourceusageestimatedbeforeactualexecutionandnisthenumberofresources.Assumed: Predicted Resource Usage Actual Resource Usage≤
.Value
of ResourceConsumption . ismorethan1generallybecause Actual Resource Usage ismorethan Predicted Resource Usage butideallyitwillbe1whenbothareequal.Inthisresearchrk,maximum values for ResourceConsumption has been fixed and that is called threshold value.Followingformulaisusedtocalculatenumberofrequestsmissed Requests
Missed( ) inaparticularperiodoftime(Equation2):
RequestsMissed
= [Number of Requests Executed Successfully – Number of Requests Missed Deadline] (2)
Forsuccessfulexecutionofresources,valueofRequestsMissed
islesserthanthresholdvalue.Algorithm1isusedtoanalysestheperformanceofmanagementofresources.
Withthehelpof(Equation1)and(Equation2),resourceconsumptioniscalculatedandallocatestheresourcesforexecutionandthencomparestheresourceconsumptionwiththresholdvalue Th
c( ) .IfresourceconsumptionislessthanthresholdvalueandvalueofRequests
Missedislessthanthreshold
value Thm( ) thenexecutionofresourcescontinuesotherwisenoresourceisallocatedandprocess
ofreallocationisstartedusingAlgorithm1.Aftermeetingthiscondition,resourcesareallocatedfor
Algorithm 1. Analyzing and Panning Unit (AU)
Journal of Organizational and End User ComputingVolume 29 • Issue 4 • October-December 2017
11
furtherexecutionandvalueofresourceconsumptionand RequestsMissed
arecheckedperiodically.Incaseofmorevaluethanthreshold,alertwillbegeneratedbyperformancemonitor.
4.4. ExecutorExecutorimplementstheplanafteranalyzingcompletely.Toreducetheexecutiontimeandexecutioncostandimproveresourceutilizationisamainobjectiveofexecutor.Basedontheoutputgivenbyanalysisandexecutortracksthenewuserrequestsubmissionandresourceaddition,andtaketheactionaccordingtorulesdescribedinknowledgebase.
4.5. EffectorEffectorisusedtoexchangeupdatedinformationanditisusedtotransferthenewpolicies,rulesandalertstoothernodeswithupdatedinformation.
5. PERFORMANCE EVALUATION
Theaimofthisperformanceevaluationistodemonstratethatitisfeasibletoimplementanddeploytheagricultureasaserviceonrealcloudresources.ToolsusedforsettingupcloudenvironmentforperformanceanalysisareMicrosoftVisualStudio2010(SaaS),Aneka(PaaS),SQLServer2008,andCitrixXenServer(IaaS).Anekahasbeeninstalledalongwithitsrequirementsonallthenodesthatprovidecloudservice.Nodesinthissystemcanbeaddedorremovedbasedontherequirement.AaaSisinstalledonmainserverandtestedonvirtualcloudenvironmentthathasbeenestablishedatCLOUDS Lab, University of Melbourne, Australia.Differentnumberofvirtualmachineshavebeeninstalledondifferentservers,anddeployedtheAaaStomeasurethevariations.Inthisexperimentalsetup,threedifferentcloudplatformsareused:SoftwareasaService(SaaS),PlatformasaService(PaaS)andInfrastructureasaService(IaaS)asshowninFigure5.
AtSaaS level,MicrosoftVisualStudioisusedtodevelope-agriculturewebservicetoprovideuserinterfaceinwhichusercanaccessservicefromanygeographicallocation.AtPaaS level,AnekacloudapplicationplatformisusedasascalablecloudmiddlewaretomakeinteractionbetweenIaaS
Figure 5. Deployment of components at runtime and their interaction
Journal of Organizational and End User ComputingVolume 29 • Issue 4 • October-December 2017
12
andSaaS,andcontinuallymonitor theperformanceof thesystem.AtIaaS level, threedifferentservers(consistofvirtualnodes)havebeencreatedthroughCitrixXenServerandSQLServerhasbeenusedfordatastorage.SchedulerasshowninFigure5,runsatIaaSlevelonCitrixXenServer.ComputingnodesusedinthisexperimentworkarefurthercategorizedintothreecategoriesasshowninTable7.Theexecutioncostiscalculatedbasedonuserrequestanddeadline(ifdeadlineistooearly(urgent)itwillbemorecostlybecausethereisaneedofgreaterprocessingspeedandfreeresourcestoprocessparticularrequestwithurgency).Thereisindividualpriceisfixed(artificially)fordifferentresourcesbecausealltheresourcesareworkingincoordinationmannertofulfillthedemandofuser(demandofuserischangingdynamically).
Experimentsetupusing3serversinwhichfurthervirtualnodes(12=6(Server1)+4(Server2)+2(Server3))arecreated.EveryvirtualnodehasdifferentnumberforExecutionComponents(ECs)toprocessuserrequestandeveryEChastheirowncost(C$/ECtimeunit(Sec)).Table1showsthecharacteristicsoftheresourcesusedandtheirExecutionComponent(EC)accesscostpertimeunitinClouddollars(C$)andaccesscostinC$ismanuallyassignedforexperimentalpurposes.TheaccesscostofanECinC$/timeunitdoesnotnecessarilyreflectthecostofexecutionwhenECshavedifferentcapabilities.TheexecutionagentneedstotranslatetheaccesscostintotheC$foreachresource.Suchtranslationhelpsinidentifyingtherelativecostofresourcesforexecutinguserrequestsonthem.Duetolimitednumberofresources,costincreaseswithincreaseinuserrequests.Costisvaryingintwodifferentcases:i)relaxeddeadlineandii)tightdeadline.Inbothcases,whenthedeadlineislow(e.g.200secs),thenumberofuserrequestsprocessedincreasesasthebudgetvalueincreases.Whenahigherbudgetisavailable,theexecutionagentusesexpensiveresourcestoprocessmoreuserrequestswithinthedeadline.Alternatively,whenschedulingwithalowbudget,thenumberofuserrequestsprocessedincreasesasthedeadlineisrelaxed.DifferentnumberofexperimentshasbeenperformedbycomparingAaaS(QoS-awareAutonomic)asdiscussedinSection 4withnon-autonomicresourcemanagementtechnique(non-autonomic)inwhichnoautonomicschedulingmechanismisconsideredwhileallocatingresourcestoprocesstheuserrequests.
5.1. DatasetsDatasets used in this research work are downloaded from the Open Government Data PlatformIndia(data.gov.in,2015),IndiaAgricultureandClimateDataSet(Sanghietal.),andregionallandandclimatemodellinginChina(Sanghietal.)canbeintheorderof1000000records,withsizeof3.5GB.Thedataiscominginlargedatavarietyandvolumefrombothusersintheformofimageslikedamagedcropimagesduetoweather,insectsetc.anddevicesthroughInternetofThings(IoT)sensorsandsatellites(GPSsystems)thatsendweatherrelatedimages.Asaresultofregularcapturingandcollectionofdatasets,theygrowwiththevelocityof80.72KB/minuteormore(Sanghietal.).FivedifferenttablesusedtoprocessthedifferenttypesofdataasdescribedinTable8toTable12.
Table 7. Configuration Details of Cloud Environment
Resource_Id Configuration Specifications Operating System
Number of Virtual Node
Number of ECs
Price (C$/EC Time
Unit)
R1 IntelCore2Duo-2.4GHz
1GBRAMand160GBHDD Windows 6 18 2
R2 IntelCorei5-2310-2.9GHz
1GBRAMand160GBHDD Linux 4 12 3
R3 IntelXEONE52407-2.2GHz
2GBRAMand320GBHDD Linux 2 6 4
Journal of Organizational and End User ComputingVolume 29 • Issue 4 • October-December 2017
13
Table 8. Crop Information
CropId Crop Name
Crop Type Soil Texture Min
LandGrowing Period
Seed Type Price Quantity
C1 Rice Kharif SlityClay 5Acre 3Months Wet 1200Rs./Kg 2Kg/Acre
C2 Maize Rabi SlityLoamClay 4Acre 4Months Dry 1600Rs./Kg 1Kg/Acre
C3 Wheat Zaid LoamClay 3Acre 3Months Wet 1000Rs./Kg 2Kg/Acre
C4 Sugarcane Cash Slity 4Acre 6Months Dry 800Rs./Kg 6Kg/Acre
Table 9. Weather information
Crop Name Temperature Season Pressure (CFM) Wind Speed Rainfall Location
Rice 15-18°C Winter 0.75to1.5 16Km/h 300–650mm Ambala
Maize 17-22°C Summer 0.05to0.5 12Km/h 100–150mm Amritsar
Wheat 25-30°C Rainy 1.5to5.2 17.3Km/h 200–250mm GangaNagar
Sugarcane 35-40°C Summer 1to10 8Km/h 400–600mm Pathankot
Table 10. Soil information
Soil Texture Bulk Density Inorganic
MaterialOrganic Material Water Air Color Structure Infiltration
SlityClay 2.60to2.75gramspercm3 Sandandclay
Plantandanimalresidues
25% 28% Brown Plate-like 15mm/hour
SlityLoamClay
2.7to2.75gramspercm3 SandandSlit Animal
residues 22% 18% Red Prism-like 10mm/hour
LoamClay 2.60to2.75gramspercm3 ClayandSlit Plantresidues 37% 21% Brown Blocklike 18mm/hour
Slity 2.60to2.75gramspercm3
Sand,SlitandClay
Plantandanimalresidues
31% 29% Black Spherelike 22mm/hour
Table 11. Pest information
Crop Type
Crop Disease Effect Treatment Pesticide Name Solubility
in Water Price Outcome
Kharif Bacterialbrownspot
Degradesoilfertility
ReduceIrrigation Carbonate Yes Rs.
1500/LImproveProductivity
Rabi Zonateeyespot
Degradeproductivity
DistributeSoil Organophosphate No Rs.
2200/LImprovesoilfertilization
Zaid Dwarfbunt
Increaseriskofotherdisease
Sprayirrigation Parathyroid Yes Rs.
2300/LReduceriskofotherdiseases
Cash Ergot Degradeproductivity
DripIrrigation Parathyroid Yes Rs.
1800/LReduceproductivity
Journal of Organizational and End User ComputingVolume 29 • Issue 4 • October-December 2017
14
5.2. Performance MetricsThefollowingmetricsareusedtocalculatetheexecutioncost,executiontime,resourceutilization,latency,detectionrateandscalabilityforprocessinguserrequestsastakenfrompreviouswork(SinghandChana,2015;Singhetal.,2015;SinghandChana,2016):
Execution TimeisaratioofdifferenceofrequestfinishtimeWFi( ) andrequeststarttimeWStart
i( ) tonumberofrequests.FollowingformulaisusedtocalculateExecutionTime(ET)(Equation3):
ETWF WStart
nii
ni i=−
=∑
1
(3)
wheren isthenumberofrequeststobeexecuted.Execution Costisdefinedasthetotalamountofcostspentperonehourfortheexecutionof
requestandmeasuredinCloudDollars(C$).Followingformulaisusedtocalculateexecutioncost(C)(Equation4):
C ET Pricei
= × (4)
Latencyisadefinedasadifferenceoftimeofinputcloudworkloadandtimeofoutputproducedwithrespecttothatworkload.FollowingformulaisusedtocalculateLatency(Equation5):
Latency timeof output producedafterexecution timei
i
n
= −=∑1
� � � � � �� � � � �of inputof cloudworkload( ) (5)
wherenisnumberofworkloads.Resource Utilizationisdefinedasaratioofactualtimespentbyresourcetoexecuteworkload
to total uptime of resource for single resource. Following formula is used to calculate resourceutilization(Equation6):
ResourceUtilizationactual timespentbyresourcet
ii
n
�� � � � �
==∑1
ooexecuteworkload
total uptimeof resource
� �
� � �
(6)
wherenisnumberofworkloads.
Table 12. Fertilizer information
Crop Type Fertilizer Name Nutrient Composition Price
Kharif Urea Nitrogeninformofurea(amide)(N) 7000Rs./10Kg
Rabi Ammonium-Nitrate AmmoniacalNitrogen,NitrogenNitrateandUreaNitrogen 9100Rs./10Kg
Zaid Ammonium-Sulphate AmmoniacalnitrogenandSulpher 6200Rs./10Kg
Cash Urea-Ammonium AmmoniacalnitrogenandNeutralammoniumcitrateSolublephosphate 13200Rs./10Kg
Journal of Organizational and End User ComputingVolume 29 • Issue 4 • October-December 2017
15
Securityismeasuredintermsofdetectionrate.Experimenthasbeenconductedwithdifferenttypeofattacks(DoS,R2L,U2RandProbing)anddifferenttoolsusedtolaunchdifferentattacksaremetasploitframeworkforDoS,HydraforR2L,NetCatforL2RandNMAPforprobing.Detection Rateistheratiooftotalnumberoftruepositivestothetotalnumberofintrusions(Sorensenetal.,2010):
DetectionRateTotal NumberofTruePositives
Total Numbe
=
rrof Intrusions (7)
Scalabilityismeasuredintermsofthroughput.Itistheratiooftotalnumberofworkloadstothetotalamountoftimerequiredtoexecutetheworkloads.Followingformulaisusedtocalculatethroughput(Equation8):
ThroughputW
Totalamountofn�
Total�Number�of�Workloads�=
( )� � �ttimerequired toexecutetheworkloads W
n� � � � � �( )
(8)
Experimenthasbeenconductedwith180userrequestsforverificationofexecutioncost,executiontime,resourceutilization,latency,detectionrateandscalability.Withincreasingthenumberofuserrequests,thevalueoflatencyisincreasing.ThevalueoflatencyinQoS-awareautonomicsystemislesserascomparedtonon-autonomicbasedresourceschedulingatdifferentnumberofuserrequestsasshowninFigure6.Themaximumvalueoflatencyis193secondsandminimumvalueoflatencyis59secondsinQoS-awareautonomicresourcemanagementtechnique.AveragelatencyinQoS-awareautonomicis15.22%lesserthannon-autonomicresourcemanagementtechnique.Thevalueofaveragecost forbothQoS-awarecloudbasedautonomic resourcemanagement techniqueand
Figure 6. Effect of change in number of user requests on latency
5.3. Experimental Results -- Based on Modelling and Simulation using CloudSim
Journal of Organizational and End User ComputingVolume 29 • Issue 4 • October-December 2017
16
non-autonomicresourcemanagementiscalculatedwithdifferentnumberofuserrequestsasshowninFigure7.Averagecostisincreasingwithincreaseinnumberofuserrequests.At180userrequests,averagecostinQoS-awareautonomicis25.36%lesserthannon-autonomicresourcemanagementtechnique.QoS-awareautonomicperformsexcellentwithdifferentnumberofuserrequests.ExecutioncostinQoS-awareautonomicis27.65%lesserthannon-autonomicresourcemanagementtechnique.
AsshowninFigure8,theexecutiontimeisincreasingwithincreaseinnumberofuserrequests.At 90 user requests, execution time in QoS-aware autonomic resource management techniqueis 24.66% lesser than non-autonomic resource management technique. After 120 user requests,executiontimeincreasesabruptlyinnon-autonomicresourcemanagementtechniquebutQoS-awareautonomicperformsbetter thannon-autonomic technique.Averageexecution time inQoS-awareautonomicis18.960%lesserthannon-autonomicresourcemanagementtechnique.Withincreasingthenumberofuserrequests,thepercentageofresourceutilizationisincreasing.ThepercentageofresourceutilizationinQoS-awareautonomicresourcemanagementtechniqueismoreascomparedtonon-autonomicresourcemanagement(non-autonomic)atdifferentnumberofuserrequestsasshowninFigure9.Themaximumpercentageofresourceutilizationis94.66%at180userrequestsinQoS-awareautonomicbutQoS-awareautonomicperformsbetterthannon-autonomictechnique.AverageresourceutilizationinQoS-awareautonomicis31.96%morethannon-autonomicresourcemanagementtechnique.
Scalability ismeasured in termsof throughput.Numberof software,networkandhardwarefaults(faultpercentage)hasbeeninjectedtoverifythethroughputoftheproposedsystemwith100userrequests.Figure10showsthecomparisonofthroughputofbothQoS-awareautonomicresourcemanagementapproachandnon-QoSbasedresourcemanagementtechnique(non-autonomic)at100userrequestsanditisclearlyshownthatQoS-awareautonomicperformsbetterthannon-autonomic.Inthisexperiment,ithasbeenfoundthemaximumvalueofthroughputatfaultpercentage45%i.e.QoS-awareautonomichas26%morethroughputthannon-autonomic.Detectionrateincreaseswithrespecttotimeanditconsidersthenumberofblockedanddetectedattacks.Fornewattackorintrusiondetection,databaseisupdatedwithnewsignaturesandnewpolicesandrulesaregeneratedtoavoid
Figure 7. Effect of change in number of user requests on execution cost
Volume 29 • Issue 4 • October-December 2017
17
sameattack.Experimenthasbeenconductedforknownattacks;itisclearlyshowninFigure11thatQoS-awareautonomicperformsbetterthansnortanomalydetector(non-autonomic).Furthersignaturesofsomeknownattackshavebeenremovedfromdatabasetoverifytheworkingofproposedsystem.
Table13describesthecomparisonofexecutiontimeusedtoprocessdifferentnumberofworkloads(90and180)oncloudenvironmentforproposedsystemwithdifferentnumberof
Figure 8. Effect of execution time with change in number of user requests
Figure 9. Effect of change in number of user requests on resource utilization
Journal of Organizational and End User Computing
Journal of Organizational and End User ComputingVolume 29 • Issue 4 • October-December 2017
18
VirtualMachines(VMs).ThenumberofVMsusedtoexecutetheworkloadswasincrementedgraduallyshowinghowthetotalexecutiontimewasreducedwhenmoreVMswereaddedtothecloud.WithonevirtualnoderunningonServerR1,executionof45workloadsfinishedin436.12seconds.With12virtualnodes(6runningonR1,4runningonR2and2runningonR3),theapplicationtook276.16seconds.Itisnotedthattheexecutiontimeisreducedwithaddingadditionalvirtualnodes.
Figure 10. Throughput [100 user requests] vs. Fault percentage (%)
Figure 11. Detection rate vs. Attacks
Journal of Organizational and End User ComputingVolume 29 • Issue 4 • October-December 2017
19
5.4. Statistical Analysis
StatisticalsignificanceoftheresultshasbeenanalyzedbyCoefficientofVariation Coff ofVar. .( ) ,astatisticalmethod.Coff ofVar. . isstatisticalmeasureofthedistributionofdataaboutthemeanvalue.Coff of Var. . isusedtocomparetodifferentmeansandfurthermoreofferanoverallanalysisofperformanceofthetechniqueusedforcreatingthestatistics.Itstatesthedeviationofthedataasaproportionofitsaveragevalue,andiscalculatedasfollows(Equation9):
Coff ofVarSD
. . �M
= ×100 (9)
whereSD isastandarddeviationandM ismean.Coff ofVar. . ofexecutiontimeandhavebeenstudied of QoS-aware autonomic resource management technique and non-autonomic resource
Table 13. Total execution time of a bulk of cloud workloads distributed in three servers
Number of WorkloadsVirtual Nodes
Total Workers Execution Time (Seconds)R1 R2 R3
45 1 0 0 1 436.12
45 1 1 0 2 428.69
45 2 1 0 3 418.97
45 2 2 0 4 407.55
45 3 2 0 5 398.17
45 4 2 0 6 380.30
45 4 2 1 7 361.66
45 4 3 1 8 345.18
45 5 3 1 9 331.21
45 5 3 2 10 315.03
45 5 4 2 11 299.97
45 6 4 2 12 276.16
90 1 0 0 1 1803.11
90 1 1 0 2 1771.18
90 2 1 0 3 1759.66
90 2 2 0 4 1736.15
90 3 2 0 5 1691.77
90 4 2 0 6 1668.96
90 4 2 1 7 1636.11
90 4 3 1 8 1625.19
90 5 3 1 9 1578.21
90 5 3 2 10 1551.68
90 5 4 2 11 1529.11
90 6 4 2 12 1503.11
Journal of Organizational and End User ComputingVolume 29 • Issue 4 • October-December 2017
20
managementtechniqueasshowninFigure12andFigure13.RangeofCoff ofVar. . (0.25%-1.69%)for execution timeand (0.37% -1.96%) for cost approves the stabilityofQoS-aware autonomicresourcemanagementtechniqueasshowninFigure12andFigure13.SmallvalueofCoff ofVar. . signifies QoS-aware autonomic resource management technique is more efficient in resourceschedulinginthesituationswherethenumberofuserrequestshaschanged.ValueofCoff ofVar. . decreasesasthenumberofuserrequestsisincreasing.
6. CONCLUSION AND FUTURE DIRECTIONS
Cloud-basedautonomicinformationsystem(AaaS)foragricultureservicehasbeenpresented,whichmanagesthevarioustypesofagriculture-relateddatabasedondifferentdomainsthroughdifferentuserpreconfigureddevices.K-NN(k-NearestNeighbor)classificationmechanismisusedtoclassifytheagriculturedata.Further,classifieddataisinterpretedanduserscaneasilydiagnosetheagriculturestatusautomaticallythroughAaaS.Inaddition,AaaSusestworesourceschedulingpolices(timeandcost)forefficientresourceallocationatinfrastructurelevelafteridentificationofQoSrequirementsofuserrequest.Theperformanceofproposedsystemhasbeenevaluatedincloudenvironmentand
Figure 12. CoV for execution time with each scheduling algorithm
Figure 13. CoV for execution cost with each scheduling algorithm
Journal of Organizational and End User ComputingVolume 29 • Issue 4 • October-December 2017
21
experimentalresultsshowthattheproposedsystemperformsbetterintermsofexecutiontime,cost,resourceutilization,latency,scalabilityandsecurity.Infuture,theproposedtechniquecanbeextendedbyincorporatingotherQoSparameterslikenetworkbandwidth,availability,customersatisfaction,computingcapacityetc.Proposedtechniquecanbeextendedbydevelopingpluggablescheduler,inwhichresourceschedulingcanbechangedeasilybasedontherequirements.
ACKNOwLEDGMENT
Oneoftheauthors,Dr.SukhpalSinghGill[PostDoctorateFellow],gratefullyacknowledgestheCLOUDS Lab, School of Computing and Information Systems, The University of Melbourne,Australia,forawardinghimtheFellowshiptocarryoutthisresearchwork.
Journal of Organizational and End User ComputingVolume 29 • Issue 4 • October-December 2017
22
REFERENCES
AgricultureData,GovernmentofIndia.(n.d.).Retrievedfromhttps://data.gov.in/catalogs/sector/Agriculture-9212
Elsheikh,R.,MohamedShariff,A.R.B.,Amiri,F.,Ahmad,N.B.,Balasundram,S.K.,&Soom,M.A.M.(2013).AgricultureLandSuitabilityEvaluator(ALSE):Adecisionandplanningsupporttoolfortropicalandsubtropicalcrops.Computers and Electronics in Agriculture,93,98–110.doi:10.1016/j.compag.2013.02.003
Hu,Y.,Quan,Z.,&Yao,Y.(2004).Web-basedAgriculturalSupportSystems.Proceeding of the Workshop on Web-based Support Systems(pp.75-80).
IndiaAgricultureAndClimateDataSet.(n.d.).Retrievedfromhttps://ipl.econ.duke.edu/dthomas/dev_data/datafiles/india_agric_climate.htm
Jeong,S.,Jeong,H.,Kim,H.,&Yoe,H.(2013).CloudComputingbasedLivestockMonitoringandDiseaseForecastingSystem.International Journal of Smart Home,7(6),313–320.doi:10.14257/ijsh.2013.7.6.30
Narayana Reddy, M., & Rao, N. H. (1995). GIS Based Decision Support Systems in Agriculture. NationalAcademyofAgriculturalResearchManagementRajendranagar.
Nikkilä, R., Seilonen, I., & Koskinen, K. (2010). Software architecture for farm management informationsystems inprecisionagriculture.Computers and Electronics in Agriculture,70(2), 328–336.doi:10.1016/j.compag.2009.08.013
Prasad,S.,Peddoju,S.K.,&Ghosh,D.(2013).AgroMobile:ACloud-BasedFrameworkforAgriculturistsonMobilePlatform. International Journal of Advanced Science and Technology,59,41–52.doi:10.14257/ijast.2013.59.04
Ruixue,Z.(2002).StudyonWeb-basedAgriculturalInformationSystemDevelopmentMethod.Proceedings of the Third Asian Conference for Information Technology in Agriculture,China(pp.601-605).
Shangguan,W.,Dai,Y.,Liu,B.,Ye,A.,&Yuan,H.(2012,February29).Asoilparticle-sizedistributiondatasetforregionallandandclimatemodellinginChina.Geoderma,171,85–91.doi:10.1016/j.geoderma.2011.01.013
Singh,S.,&Chana,I.(2015).QoS-awareAutonomicResourceManagementinCloudComputing:ASystematicReview.ACM Computing Surveys,48(3),1–46.doi:10.1145/2843889
Singh,S.,&Chana, I. (2015).QRSF:QoS-aware resourcescheduling framework incloudcomputing.The Journal of Supercomputing,71(1),241–292.doi:10.1007/s11227-014-1295-6
Singh,S.,&Chana,I.(2015).Q-aware:Qualityofservicebasedcloudresourceprovisioning.Computers & Electrical Engineering,47,138–160.doi:10.1016/j.compeleceng.2015.02.003
Singh,S.,&Chana,I.(2016).EARTH:Energy-awareAutonomicResourceSchedulinginCloudComputing.Journal of Intelligent and Fuzzy Systems,30(3),1581–1600.doi:10.3233/IFS-151866
Sørensen,C.G.,Fountas,S.,Nash,E.,Pesonen,L.,Bochtis,D.,Pedersen,S.M.,&Blackmore,S.B.et al.(2010).Conceptualmodelofafuturefarmmanagement informationsystem.Computers and Electronics in Agriculture,72(1),37–47.doi:10.1016/j.compag.2010.02.003
Sørensen,C.G.,Pesonen,L.,Bochtis,D.D.,Vougioukas,S.G.,&Suomi,P.(2011).Functionalrequirementsforafuturefarmmanagementinformationsystem.Computers and Electronics in Agriculture,76(2),266–276.doi:10.1016/j.compag.2011.02.005
Journal of Organizational and End User ComputingVolume 29 • Issue 4 • October-December 2017
23
Sukhpal Singh Gill joined Computer Science and Engineering Department of Thapar University, Patiala, India, in 2016 as a Faculty. Presently, Dr. Gill is working as Post Doctorate Fellow at CLOUDS Lab, School of Computing and Information Systems, The University of Melbourne, Australia. Dr. Gill obtained the Degree of Master of Engineering in Software Engineering from Thapar University, as well as a Doctoral Degree specialization in “Autonomic Cloud Computing” from Thapar University. Dr. Gill received the Gold Medal in Master of Engineering in Software Engineering. Dr. Gill is a DST Inspire Fellow [2013-2016] and worked as a SRF-Professional on DST Project, Government of India. He has done certifications in Cloud Computing Fundamentals, including Introduction to Cloud Computing and Aneka Platform (US Patented) by ManjraSoft Pty Ltd, Australia and Certification of Rational Software Architect (RSA) by IBM India. His research interests include Software Engineering, Cloud Computing, Internet of Things and Fog Computing. He has more than 40 research publications in reputed journals and conferences.
Inderveer Chana joined Computer Science and Engineering Department of Thapar University, Patiala, India, in 1997 as Lecturer and is presently serving as Professor in the department. She is Ph.D. in Computer Science with specialization in Grid Computing, M.E. in Software Engineering from Thapar University and B.E. in Computer Science and Engineering. Her research interests include Grid and Cloud computing and other areas of interest are Software Engineering and Software Project Management. She has more than 100 research publications in reputed Journals and Conferences. Under her supervision, more than 40 ME thesis and seven Ph.D thesis have been awarded and five Ph.D. thesis are on-going. She is also working on various research projects funded by Government of India.
Rajkumar Buyya is a Fellow of IEEE, Professor of Computer Science and Software Engineering and Director of the Cloud Computing and Distributed Systems (CLOUDS) Laboratory at the University of Melbourne, Australia. He is also serving as the founding CEO of Manjrasoft, a spin-off company of the University, commercialising its innovations in Cloud Computing. He has authored over 500 publications and four text books. He is one of the highly cited authors in computer science and software engineering worldwide (h-index 110+, 60000+ citations). He has served as the founding Editor-in-Chief (EiC) of IEEE Transactions on Cloud Computing and now serving as Co-EiC of Journal of Software: Practice and Experience.
top related