A Method for Validating Intent Model Behavior in DSVMs

Abstract

The direct runtime interpretation and execution of domain-specific models through the use of a Domain Specific Virtual Machine (DSVM) is an area of emerging relevance in the model-driven engineering community. This is due in part to the increased efficiency and decreased complexity achieved through specialization of the architecture in disparate domains. An approach to the design of a DSVM is to include a middleware that is responsible for the delivery and management of domain-specific services. It is the job of this middleware to help realize user intent through the execution of received commands while ensuring adherence to system policies based on changing environmental context. To provide assurance of functionality, the DSVM middleware must be policy and context-aware and facilitate variability in its operations. It achieves this variability by dynamically generating behavioral models for execution in response to commands. The dynamic generation of models poses the challenge of ensuring their correctness at runtime. To guarantee the correctness of generated models, we adopted model validation techniques to ensure policy compliance and employed the Alloy Analyzer in our prototype to demonstrate the efficacy of this approach. This granted us use of the Alloy specification language, which, by utilizing first-order logic, enhanced our model validation process by allowing more expressive policies. We demonstrate the increased capabilities and assurance realized by this approach through a case study with a DSVM middleware instance for the communication domain.