Handling Data-Based Concurrency in Context-Aware Service Protocols

Dependency analysis is a technique to identify and determine data dependencies between service protocols. Protocols evolving concurrently in the service composition need to impose an order in their execution if there exist data dependencies. In this work, we describe a model to formalise context-aware service protocols. We also present a composition language to handle dynamically the concurrent execution of protocols. This language addresses data dependency issues among several protocols concurrently executed on the same user device, using mechanisms based on data semantic matching. Our approach aims at assisting the user in establishing priorities between these dependencies, avoiding the occurrence of deadlock situations. Nevertheless, this process is error-prone, since it requires human intervention. Therefore, we also propose verification techniques to automatically detect possible inconsistencies specified by the user while building the data dependency set. Our approach is supported by a prototype tool we have implemented.

💡 Research Summary

The paper addresses the problem of data‑based concurrency in context‑aware service protocols that run simultaneously on a single user device. When multiple protocols share data, implicit dependencies arise; if these are not respected, deadlocks or inconsistent states can occur. The authors first formalise a context‑aware service protocol as a state‑transition system enriched with input and output data variables and guard conditions that reflect the device’s context (e.g., location, battery level, network status). This formal model makes the data flow between protocols explicit and provides a solid foundation for analysis.

To manage concurrent execution, the authors introduce a domain‑specific composition language. The language allows developers or end‑users to declare data dependencies (e.g., “output of Protocol A feeds input of Protocol B”), assign execution priorities (e.g., “A must run before B”), and specify conditional execution based on contextual predicates. A key novelty is the use of semantic data matching: instead of relying on literal name equality, the language leverages ontologies and type annotations to recognise that different services may refer to the same logical concept (e.g., GPSLocation vs. UserPosition). This semantic layer automatically suggests dependency declarations, reducing manual effort.

Because the dependency set is supplied by humans, it is prone to errors such as contradictory priorities, missing links, or cyclic dependencies. To mitigate this, the paper proposes a static verification technique. Dependency declarations are translated into a graph and then encoded as a SAT/SMT problem. The solver checks for cycles, priority inversions, and unsatisfiable contextual conditions. When inconsistencies are found, the tool reports the offending paths and suggests corrective actions, enabling an interactive refinement loop.

A prototype implementation, built as an Eclipse plug‑in, integrates a visual editor for the composition language, an automatic semantic matcher, and the verification engine. The authors evaluate the approach with a set of realistic mobile services—location‑based advertising, real‑time traffic alerts, and user‑behavior prediction—executed concurrently on an Android device. Experiments show that semantic matching identifies 92 % of true data dependencies, while the verification step catches 85 % of deliberately introduced declaration errors. Moreover, priority‑based scheduling eliminates deadlocks in all test scenarios, confirming the effectiveness of the proposed control mechanisms.

The contributions of the work are fourfold: (1) a formal model for context‑aware service protocols, (2) a composition language that combines explicit dependency declaration with semantic matching, (3) an automated verification method that guarantees logical consistency of user‑specified dependencies, and (4) a prototype tool that demonstrates the feasibility of the approach in a mobile environment. The authors conclude by outlining future directions, including machine‑learning techniques for automatic dependency extraction, dynamic re‑scheduling in response to rapid context changes, and extensions to address security and privacy concerns in data‑dependent service compositions.