Section: New Results

Service-oriented Computing in the Future Internet

Participants : Georgios Bouloukakis, Nikolaos Georgantas, Valérie Issarny, Ajay Kattepur.

With an increasing number of services and devices interacting in a decentralized manner, choreographies represent a scalable framework for the Future Internet. The service oriented architecture inherent to choreographies allows abstracting multiple devices as components, that interact through middleware connectors via standard protocols. However, the heterogeneous nature of devices leads to choreographies that not only include conventional services, but also sensor-actuator networks, databases and service feeds. We reason about their behavior through abstract middleware interaction paradigms, such as client-service (CS), publish-subscribe (PS) and tuple space (TS), made interoperable through the eXtensible Service Bus (XSB) connector.

Extensible Service Bus for the Future Internet: XSB is an abstract service bus that deals effectively with the cross-integration of heterogeneous interaction paradigms [17] . Inside the XSB, the CS, PS and TS paradigms are modeled as abstract base connectors. Their space coupling semantics are represented with programming interfaces used by applications (APIs) and corresponding application interface description languages (IDLs). Their behavioral semantics are formally specified in terms of LTS (Labeled Transition Systems). We formally verify the correctness of these behavioral specifications with respect to time coupling and concurrency properties expressed in LTL temporal logic. This allows stating the correctness of the connector models with respect to the semantics that they must have. This further enables identifying the behavioral semantics of the XSB connector derived from the interconnection of base connectors. More specifically, in order to identify the time coupling and concurrency semantics of XSB and construct a converter among the base connectors, we build upon the formal method of protocol conversion via projections (Lam, S.S.: Protocol Conversion. IEEE Trans. Softw. Eng. 14(3) (1988) 353–362.). According to this method, conversion between two different protocols is possible if both protocols can be projected (where projection is an abstraction defined as a set of transformations on the protocol LTS) to a functionally sufficient common image protocol. Then, the end-to-end protocol of the interconnection of the two protocols is this image protocol.

We have implemented our XSB solution into an extensible development and execution platform for application and middleware designers. Using this platform, they can easily develop composite applications: they only need to build descriptions for the constituent services and directives for data mapping among them. Our platform then deals with reconciling among the heterogeneous interaction paradigms and protocols of the services by employing binding components (BCs) that adapt between the native middleware of the services and the XSB bus protocol. The XSB itself is implemented on top of an existing ESB substrate. Support for new middleware platforms, new ESB substrates, or even new interaction paradigms can be incorporated in a facilitated way thanks to the provided XSB architectural framework.

QoS composition and analysis of heterogeneous choreographies: Leveraging on the functional interoperability across interaction paradigms offered by the XSB, we study the Quality of Service (QoS) performance of choreographies [21] . QoS dependency plays an important role in the service oriented system lifecycle, including discovery, runtime selection, replacement and contractual guarantees. Consequently, QoS composition among choreographed devices should tackle multi-dimensional probabilistic metrics combined with message passing constraints imposed at design-time. We make use of an algebraic QoS composition model that is applied at the interaction paradigm level to study the composition of QoS metrics, and the subsequent tradeoffs. While traditional QoS composition analysis has been done purely at the application level, analyzing the effect of middleware interactions allows us to study CS, PS and TS based device compositions. This produces interesting insights such as selection of a particular system and its middleware during design-time, or end-to-end QoS expectation/guarantees during runtime. Our formulation also allows for runtime reconfiguration, in order to optimally produce design time QoS expectations. Such flexible reconfiguration policies are crucial in the case of large scale choreographies with high variability in runtime performance of participating devices.

Further, we study the effect of time/space coupling on the latency of successful transactions across the XSB connector [20] . XSB models the message passing among peers through generic post and get operations, that represent peer behavior with both tight (CS) and loose (PS/TS) time/space coupling. The heterogeneous lease and timeout behaviors of these operations severely affect latency and success rates of messages passed either synchronously or through callbacks. By precisely studying the timing thresholds using timed automata models, we verify conditions for accurate message transactions with XSB connectors. This offers choreography designers the ability to set these timing thresholds (bottom-up) or select a particular interaction paradigm (top-down) for runtime enaction.