Section: Software

Emergent Middleware Enablers

Participant : Valérie Issarny [correspondent] .

As part of our research work on Emergent MIddleware, we have implemented Enablers (or Enabler functionalities) that make part of the overall Connect  [30] architecture realizing Emergent Middleware in practice.

Discovery Enabler: The Connect Discovery Enabler is the component of the overall Connect architecture that handles discovery of networked systems (NSs), stores their descriptions (NS models), and performs an initial phase of matchmaking to determine which pairs of systems are likely to be able to interoperate. Such pairs are then passed to the Synthesis Enabler so that mediators can be generated. The Discovery Enabler is written in Java and implements several legacy discovery protocols including DPWS and UPnP. Matchmaking is done on the basis of affordances contained in the system description, that is, ontological concepts describing the system's category. Systems with the same affordance, or affordances standing in a specialization relationship, can be considered for connection. If a system does not provide its affordance, the Discovery Enabler can infer a likely one using text categorization based on the system's interface description. The Discovery Enabler will soon be available for download under an open source license.

Synthesis Enabler: We have implemented (in Java) two approaches to mediator synthesis as part of the Connect Synthesis Enabler:

• Mapping-based mediator synthesis. This implementation focuses on networked systems that have compatible functionalities but are unable to interact successfully due to mismatching interfaces and/or behaviors. The ontology used in our implementation is encoded so as to make the reasoning more efficient at runtime while considering both subsumption and the union of classes. Based on the interface mapping, a correct-by-construction mediator is generated. In our current implementation, we are leveraging the LTSA (Labeled Transition System Analyser) model checker to generate the parallel composition of the mapping processes and verify that the overall system successfully terminates. In the near future, we will be incorporating our techniques so as to deal with ambiguous mappings, i.e., when an action from one system may semantically be mapped to different actions from the other system.

• Goal-based abstract mediator synthesis. This implementation considers the protocols of two networked systems and produces the mediator protocol that allow them to interact so as to satisfy user goals. More specifically, the alphabet of the two protocols are aligned using ontology matching. The aligned protocols as well as the user goal are encoded as a satisfiability problem. The Zot model checker solves this problem (if possible) and produces a possible feasible interaction satisfying user goals.

The synthesis enabler will soon be available for download under an open source license.