Hybrid Composition of Microservices: A Metrics-based Analysis

Abstract

"Microservices" is an architectural and organizational style in software design and development in which there is a composition mechanism for independent microservices to call, communicate, and message each other within an application. The microservices composition approach makes design easier to scale, faster to develop, and can accelerate the introduction of new features into applications. To satisfy business requirements, selecting the proper composition style is important for software development; otherwise, application development may fail. The objective of this research is to investigate the hybrid method for composing microservices and compare it with other composition approaches (choreography and orchestration), using quality metrics from the software engineering and business process modeling literature. More precisely, we make use of coupling, cohesion, and scalability metrics to analyze BPMN models representing e-commerce scenarios modeled as microservice compositions. This thesis follows the five steps: research problem identification and objectives, requirement analysis and system design, model design and development, model testing and deployment, and evaluate our BPMN models representing microservice compositions. We develop multiple BPMN workflows as artifacts to analyze choreography, orchestration, and hybrid styles for the microservices composition of e-commerce scenarios. We propose several hybrid models by integrating orchestrations and choreographies in the same workflow. We created a series of small, mid-sized, and end-to-end workflows of e-commerce scenarios. At the tool level, we use the Camunda Modeler, Camunda Platform 8 (as the automation process engine), and Amazon Web Services (AWS) to design, develop, and deploy our models. Finally, we use our calculations of the coupling, cohesion, and scalability measures to reveal an understanding of modeling microservices choreography, orchestration, and hybrid approaches, and we discuss when to use a specific approach for microservices composition. We have found from the evaluation that our proposed models are less tightly coupled compared to those modeled using orchestration and choreography. However, we also discovered that the orchestration style offers better scalability and a lower ratio of coupling and cohesion compared to choreography and hybrid approaches.