Microservice Ambients: An Architectural Meta-modelling Approach for Microservice Granularity

Isolating fine-grained business functionalities by boundaries into entities called microservices is a core activity underlying microservitization. We define microservitization as the paradigm shift towards microservices. Determining the optimal microservice boundaries (i.e. microservice granularity) is among the key microservitization design decisions that influence the Quality of Service (QoS) of the microservice application at runtime. In this paper, we provide an architecturecentric approach to model this decision problem. We build on ambients - a modelling approach that can explicitly capture functional boundaries and their adaptation. We extend the aspect-oriented architectural meta-modelling approach of ambients - AMBIENT-PRISMA - with microservice ambients. A microservice ambient is a modelling concept that treats microservice boundaries as an adaptable first-class entity. We use a hypothetical online movie subscription-based system to capture a microservitization scenario using our aspect-oriented modelling approach. The results show the ability of microservice ambients to express the functional boundary of a microservice, the concerns of each boundary, the relationships across boundaries and the adaptations of these boundaries. Additionally, we evaluate the expressiveness and effectiveness of microservice ambients using criteria from Architecture Description Language (ADL) classification frameworks since microservice ambients essentially support architecture description for microservices. The evaluation focuses on the fundamental modelling constructs of microservice ambients and how they support microservitization properties such as utility-driven design, tool heterogeneity and decentralised governance. The evaluation highlights how microservice ambients support analysis, evolution and mobility/location awareness which are significant to quality-driven microservice granularity adaptation. The evaluation is general and irrespective of the particular application domain and the business competencies in that domain.