View
398
Download
0
Category
Preview:
DESCRIPTION
A Middleware Platform to Federate ComplexEvent Processing
Citation preview
1
A Middleware Platform to FederateComplex Event Processing
Fawaz Paraïso, Gabriel Hermosillo, Romain Rouvoy, Philippe Merle, Lionel Seinturier
The Sixteenth IEEE International EDOC Conference (2012)
University of Lille & Inria lille-Nord Europe (France)
2
Agenda
• Motivation
• Challenges
• Contribution
• Validation
• Conclusion & Perspectives
3
Motivation
• What do we mean by event?
– A piece of data that represents somethinghappened in the real world
• Event-driven behaviour in daily life
– Computer
– Systems
– …
4
Motivation
• Events are everywhere
Produce events
5
Motivation
• Events are useless if they are not filtered and correlated
Events
Processing
6
Motivation
• What is Complex Event Processing (CEP)?
– Real time processing
– Intelligent business applications
• What applications can benefit from CEP?
– Real-time supply chain management
– Algorithm trading
– Monitoring (transaction, network, …)
– Credit card fraud detection
7
Motivation
• The need for real-time processing of information is relevant for many systems
– Business activity monitoring
– Fraud detection
– Nuclear crisis management
8
Motivation
Population
Experts
Localauthority
Police Firemen
EmergencyMedicalService
Nuclear Central
Media
Army
DecisionOperation
RadiationSurvey Network
NationalWeatherForecast
9
Agenda
• Motivation
• Challenges
• Contribution
• Validation
• Conclusion & Perspectives
10
Challenges
• Challenge 1: Communication heterogeneity
11
Challenges
• Challenge 2: Heterogeneous CEP Engines
CEP
Esper Etalis
StreamCruncher
ruleCore Server
12
Challenges
• Challenge 3: Scalability
Performance&
Scalability
13
Challenges
• Challenge 4: Adaptability
14
Agenda
• Motivation
• Challenges
• Contribution
• Validation
• Conclusion & Perspectives
15
Contribution
• A Middleware Platform to Federate ComplexEvent Processing
– Federate distributed CEP Engines
– Supports multiple communication services
REST, JMS, WS-Notification
– The DiCEPE Platform is an SCA-based solution
– Implemented in SCA using FraSCAti
Reflective component model
Runtime adaptative system
16
Contribution
• Distributed Platform Architecture
DiCEPE
DiCEPE
DiCEPE
17
Contribution
• Platform Architecture
DiCEPEContext
Engine Statement
*
Listener
*
BindingRest
BindingJMS
LegendComposite
Service
Property
Component
Reference
Wire
18
Contribution
• Platform Architecture
– Communication heterogeneity
Orchestrate heterogenenous services
Different bindingsREST, WS, JMS, JNA, UPnP, RPC ,RMI, JGroups, etc.
– Reconfiguration capability
Dynamic reconfigurable runtime architecture
19
Contribution
• Platform architecture
– Facililate the integration of CEP engine
Compose an heterogenous piece of software to build a new service
Supports variousImplementation technologies (Java, BPEL, C, C++, Python, …)
Interface definiton language (WSDL, Java)
20
Contribution
• The Platform adresses the challenges of :
– Communication heterogeneity
– Heterogeneous CEP
– Scalability
– Adaptability
21
Agenda
• Motivation
• Challenges
• Contribution
• Validation
• Conclusion & Perspectives
22
Validation
• DiCEPE for nuclear crisis management
Available here: http://dicepe-broker.soceda.cloudbees.net
23
Validation
• The SCA validates the challenge:
– Communication heterogeneity
– Heterogeneous CEP
– Scalability
– Adaptability
24
Validation
• Integration with the Esper and Etalis engine
Overview of Esper Engine Architecture
EsperServiceProvider
Event object
Listeners
EPLStatements
Co
nfigu
ration
1 2
3 4
5
DiCEPEArchitecture
EventExecutionWorker
PrologEngineWrapper
EtalisEventListener
InputEvents
Etalis
EtalisWrapper
PrologOutputEvents
1
2
3
statement4
1 2 3
4
5
1 2
4
3
25
Validation
• The integration of Esper and Etalis CEP engine validates the challenge :
– Communication heterogeneity
– Heterogeneous CEP
– Scalability
– Adaptability
26
Validation
• DiCEPE Cost Analysis
Implementation Avg. Exec. Time SCA overhead
Esper 27 sec -
DiCEPE (Esper+ FraSCAti) 30 sec 11%
27
Validation
• DiCEPE Scalability
Firemen Events Failures Avg. Sessions Avg. response
10,000 500,000 0 89 0.113 ms
15,000 750,000 0 135 0.142 ms
+ 50% + 50% + 51% + 26%-
28
Validation
• The scalability analysis validates the challenge
– Communication heterogeneity
– Heterogeneous CEP
– Scalability
– Adaptability
29
Validation
• Dynamic reconfiguration
30
Validation
• The FraSCAti validates the challenge:
– Communication heterogeneity
– Heterogeneous CEP
– Scalability
– Adaptability
31
Agenda
• Motivation
• Challenges
• Contribution
• Validation
• Conclusion & Perspectives
32
Conclusion & Perspectives
• DiCEPE offers interoperability between CEP engines via federation
• Flexible component architecture– Successful integration and validation of CEP engines– Multiple communication protocols
• Real scalability
• Integrate a Domain Specific Language(DSL) to express rules• Deployment of DiCEPE on heterogeneous cloud
environments• Error handling capabilities for distributed environments
33
Thank you
Questions?
@email: fawaz.paraiso@inria.fr
Recommended