Many of the modern IT systems in use today are distributed systems with a demand for high flexibility and adaptivity at runtime. Dynamic adaptations in these systems can be modelled using roles, which capture the context- and relation-dependant aspects of objects.

While role-based programming has been investigated thoroughly, existing approaches lack a native concept of concurrency and distribution. In combination with adaptivity, these concepts need to be considered carefully to ensure a valid system configuration at any time.

To address this, we propose a new meta-model for distributed role-based systems that combines the actor model of computation and the compartment role object model for structured contexts. Actors are hierarchical objects that communicate asynchronously via message passing and can thus be executed in parallel. In our approach, role-plays are grouped into transactions to ensure a consistent system state at all times.

We discuss a prototypical implementation of our approach as a programming language. Additionally, we evaluate what aspects of role-based systems are supported by our concept and show performance characteristics compared to an established role-based programming language.