The Case for Interaction Paradigm Interoperability Georgios - - PowerPoint PPT Presentation

Service-oriented Distributed Applications in the Future Internet: The Case for Interaction Paradigm Interoperability

Georgios Bouloukakis1

Joint work with Nikolaos Georgantas1, Sandrine Beauche2, Valérie Issarny1

1 Inria Paris-Rocquencourt, France 2 Ambientic, France

IFIP WG on Services-oriented Systems, Amsterdam, 11/10/2013

Service-oriented applications in the FI: Search & Rescue (S&R)

JMEDS Web service JavaSpaces data space

• 2

Web Services life sign sensors S&R personnel planning & deployment S&R coordinator monitoring & control centers JMS event-based system S&R service choreography JavaSpaces JMS

Middleware interoperability required

Info diffusion & coordination

Heterogeneous middleware interaction protocols

Interaction paradigms in middleware

• 3

JavaSpaces Client-Server (CS) Publish- Subscribe (PS) Tuple Space (TS) JMS

• Look for comprehensive systematic solution to middleware

interoperability

• Classify middleware implementations into families
• Families follow well-known interaction paradigms

Web Services Interaction paradigms Interaction protocols

publishers publishers

Some informal semantics of interaction paradigms

• 4

Space coupling

client server message(destination) publishers broker 2.event(topic) subscribers 3.event(topic) subscribers subscribers 1.subscription(topic) publishers publishers writers tspace 1.tupl e subscribers 2.tuple(template) subscribers readers/take rs 1.request-msg(destination) 2.reply-msg(source)

CS PS TS

Time coupling Concurrency

State-of-the-art in interaction paradigm interoperability

• 5

Theoretical approaches for individual paradigms

• Rely on concurrency theory, process algebras, architectural

connectors

• No study of semantics across interconnected paradigms

Practical cross-paradigm approaches

• Typically apply to specific middleware

implementations

ESB service A application middleware service B Binding Component (BC) common bus protocol

Enterprise Service Bus (ESB)

Solution to middleware interoperability

• 6

GA connector

Planning & deployment Info diffusion & coordination PS connector TS connector connector converter CS connector Sensing JavaSpaces CS PS TS Web Services Generic Application (GA) connector Interaction paradigm connectors Middleware platforms Higher-level connector

Mapping of space coupling – IDLs

• 7

CS PS TS GA element message event tuple data main scope system ID system ID system ID system ID sub-scope

• peration

topic template data qualifier interaction semantics {one-way, notification, request-response, solicit-response} {publish, subscribe} {write, take, read} {post, get, post-get, get-post} S&R element in: {sensorLocation, lifeSign}

• ut:

{personnelLocation}

• ut:

{personnelId, personnelLocation}

• ut:

{sensorLocation, lifeSign} in/out data

Time coupling and concurrency of CS, PS, TS connectors

• 8
• Specify connector protocols with

Labeled Transition Systems

PS connector

• Express and verify semantics in LTL temporal logic
• 8

fluent SUBSCRIBED = < {subscribe}, {unsubscribe} > assert EVENT_RECEIVED_IF_SUBSCRIBED_BEFORE_PUBLISH = forall [e:EVENT] [] ((SUBSCRIBED && publish[e]) -> (!event_lost[e] U get_next_return[e]))

Mapping of time coupling and concurrency

• 9

GA connector

Planning & deployment Info diffusion & coordination PS connector TS connector connector converter CS connector Sensing protocol projection

1 S. S. Lam. Protocol Conversion. IEEE Trans. Softw. Eng. 14, 3. 1988.

• Rely on the method of protocol conversion via projections1
• Common semantics of CS, PS, TS apply to GA end-to-end

XSB XSB on top of EasyESB1

GA semantics common bus protocol application middleware BC middleware CS/PS/TS GA Sensing JavaSpaces BC JavaSpaces TS GA Planning & deployment Web Services BC Web Services CS GA Info diffusion & coordination JMS BC JMS PS GA

Implementation of the GA connector: an eXtensible Service Bus (XSB)

• 10
• XSB is a generic bus
• Build on top of a substrate bus
• Cross-paradigm integration in

the S&R application

1 https://research.linagora.com/display/easyesb

XSB architectural framework

• 11

From/to substrate bus From/to middleware Core Engine Envelope for Substrate Bus GA-IDL Processor

BC extensible architecture

middleware platform level interaction paradigm level generic level Refine to introduce a new application service Refine to introduce a new middleware Refine to introduce a new interaction paradigm Refine to introduce a new substrate bus

Evaluation – Effort for S&R application design

• 12

CS-, PS-, TS-IDL descriptions (XML lines) Generated GA-IDL descriptions (XML lines) Data mapping directives (XML lines) All 3 application services 407 249 226

Development effort for composing the three available application services into the S&R application

Evaluation – Extensibility of the XSB framework

Development effort for the JMS binding component

• 13

Lines of code PS-, GA-IDL schemas (XML lines) Configuration files (XML lines) Total effort 18021 2836 248 Developer’s effort 1162 191 12 Developer’s effort ratio 6% 6% 4%

Evaluation - Performance

• 14

Interconnection Latency overhead (ms) Comparison CS-CS via EasyESB 258 – CS-CS via XSB 261,5 +1,4% CS-PS via XSB 283 +9,7% CS-TS via XSB 276 +7,0% PS-TS via XSB 298 +15,5%

• Latency overhead introduced by the bus for all interconnection

combinations

• Comparison between XSB and substrate bus

Conclusion and future perspective

• 15
• Service-oriented applications in the Future Internet
• Require cross-middleware interoperability
• Tackle this challenge via
• Abstractions and mappings for interaction paradigms
• Assessment of end-to-end semantics
• Implementation into an extensible service bus
• Next step
• Extend with support for continuous interactions – data streaming

protocols

Further information: XSB: xsb.inria.fr CHOReOS: www.choreos.eu Inria ARLES: www.rocq.inria.fr/arles

Thank you

• 16