Qualitative evaluation of dependency graph representativeness

Abstract

<p>Background: Enterprise application and open source software (OSS) platform and infrastructure projects are often today agile time-boxed projects. To enable project scaling, microservices software architecture (MSA) is considered to enable autonomous cross-functional teams. MSA results to loosely coupled services which communicate via well-designed APIs. Previous research on automated extraction of Microservice Dependency Graphs (MDGs) could provide means of reducing this documentation effort. Aims: The aim of the study was to look at the MDG representativeness of a Spinnaker OSS project micro-services-based software architecture and MDG, providing assessment of possibilities in using MDGs for documenting microservices-based software architectures. Method: The study uses a qualitative approach to evaluate the MDG representativeness of software architecture description. Evaluation is done through assessment of limitations, issues and future development possibilities. Results: MDG of Spinnaker OSS is extracted with an automation tool and contrasted to the software architecture as described on OSS project documentation. Compile-time MDG description and runtime focused documented software architecture lead to limitations in MDG rpresentativiness. Conclusions: Focusing on a particular OSS microservices project, the MDG extraction through static code analysis limits to compile-time information. Limitations in capturing inter-service communication at runtime to describe key architectural views of software architecture lead to a need to look for complementing approaches.</p>