Adaptation Consistency of Distributed Role-Oriented Applications Based on the Actor Model of Computation

Complex software is expected to adapt to changing circumstances autonomously. In software engineering, structural adaptations of objects can be modeled using contexts and roles. Existing role concepts support autonomous adaptation but provide only limited support for distributed setups. This paper addresses two major issues of distributed adaptive systems. Firstly, the mapping of roles to the actor model, which facilitates the concurrent execution of independent objects and their communication via message passing. Secondly, a concept for coordinating adaptations across multiple distributed nodes, relying on a strong or eventual consistent adaptation of distributed role-based applications. The former approach always ensures a globally consistent state of the system. The latter increases the system’s availability in the case of network congestions or failures of devices during the adaptation process by superseding a rollback of all intermediate changes. We discuss the cases for applying different consistency criteria and show the feasibility of our approach by presenting an empirical case study of a file server application in a prototypical programming language.