Page 1
Laboratory for Advanced Collaboration
LAC
AMiddlewareforData-centricandDynamicDistributedComplexEventProcessingfor
IoTReal-timeAnalyticsintheCloud
Gustavo B. Baptista, Felipe Carvalho, Sergio Colcher and Markus Endler
Department of Informatics – Pontifical Catholic University of Rio de Janeiro (PUC-Rio)
Rio de Janeiro, Brazil.{gbaptista,fcarvalho,colcher,endler}@inf.puc-rio.br
Page 2
2
Introduction
§ IoTbigdatareal-timeanalyticssystems• Massiveamountsofdata• Streamsproducedbydistributeddatasources
§ ReactiveParadigm• DistributedCEPSystemsareverysuitable
• Challenges indeployingandmanagingprocessinglogicatexecutiontime
• 24x7availability
Page 3
3
Data-centricParadigm
§ Data-centricParadigm• Themeanofinteractionisdata.(vs.message-centric,isthemessage)
• Themiddleware§ Hasthedefinitionofstructureanddata§ Awareofcontents (i.e.instances)ofstructures§ Imposesrulesonstructures, changesandaccess§ Managesdistributedstate
• Data-centricPublish-Subscribe (DCPS)§ GlobalSharedDataSpace§ Logicaldecentralized spacemaintainedbyallpeers§ Containsthestructureandinstancesofdata§ Nodesread/writedata§ Infrastructure ensuresallparticipantsaconsistentandup-to-dateview
Page 4
4
Data-CentricParadigm
§ DataDistributionServiceforReal-TimeSystems(OMG-DDS)• Fullydistributedpeer-to-peer (i.e.broker-less)• Real-timedata-centricpublish/subscribe• Highperformancecommunication,scalabilityandavailability• SpecificationofQualityofService(QoS)contracts• Mechanismsfordealingwithreal-timeaspects• PriorityandotherspecificQoSpolicies• Interoperabilityacross
§ DDSimplementations§ Programminglanguages§ Platforms
• Automaticdiscovery
Page 5
5
Data-CentricParadigm
§ DataDistribution ServiceforReal-TimeSystems(OMG-DDS)
Page 6
6
D3CEPMiddleware
§ WepresentamiddlewareforDistributedCEP• BenefitsofData-centricanddynamicdesignapproach
§ Dynamicdefinition anddeploymentofCEPrules§ Peer-to-peerroutingofeventsamongCEPrules§ Reducedcouplingofproducers,consumersandCEPrules§ Availabilityprovidedbypeer-to-peermodel§ Highthroughputandlowlatencyincommunication&detection§ EsperasaCEPengineateachnode
• Architectureandtestsregardingperformance andscalability
Page 7
7
D3CEPMiddleware
§ GlobalSharedReactiveDataSpace• Datadisseminationandreactivebehaviormodeledtogether
§ Descriptionofeventsatprocessingandcommunicationlayers• Consumers,producersandCEPrules
§ Dynamicallydefinedanddeployed seamlessly• Additionalmechanisms:
§ GlobalCatalogofeventtypes§ GlobaldefinitionofDCEPentitiesanddeployment§ EPAs,EPNs,CEPRules§ CEPservices
Page 10
10
Architecture
§ ProcessingNodeDaemon
Page 11
11
Architecture
§ MetadataSharingService
Page 12
12
Architecture
§ D3CEPAdministration
Page 13
13
Architecture
§ DynamicDDSTopicsService
Page 14
14
UseCase
§ TelemetryApplication
Page 15
15
Evaluation
§ DetectionTime
Page 16
16
Evaluation
§ DetectionTime
Page 17
17
Evaluation
§ Throughput
Page 18
18
Evaluation
§ Throughput
Page 20
20
ConclusionsandFutureWork
§ MainContributions• Data-centricdesignapproachtoDCEP
§ UseofDDSforpeer-to-peer routingofevents§ DynamicDeploymentandAutomaticDiscovery
§ FacilitatesdeploymentofCEPrules§ DistributedStateManagement
§ CEPrulesseamlesslyread/writeGlobalSharedDataSpace
§ Futurework• QoScontractsatthedetection level[Appeletal.2010]§ IoTandmissioncriticalapplications
Page 21
21
Questions
Thank you