Assessing Architecture Conformance to Coupling-Related Infrastructure-as-Code Best Practices: Metrics and Case Studies

Abstract

Infrastructure as Code (IaC) is an IT practice that facilitates the management of the underlying infrastructure as software. It enables developers or operations teams to automatically manage, monitor, and provision resources rather than organize them manually. In many industries, this practice is widespread and has already been fully adopted. However, few studies provide techniques for evaluating architectural conformance in IaC deployments and, in particular, aspects such as loose coupling. This paper focuses on coupling-related patterns and practices such as deployment strategies and the structuring of IaC elements. Many best practices are documented in gray literature sources, such as practitioner books, blogs, and public repositories. Still, there are no approaches yet to automatically check conformance with such best practices. We propose an approach based on generic, technology-independent metrics tied to typical architectural design decisions for IaC-based practices in microservice deployments to support architecting in the context of continuous delivery practices. We present three case studies based on open-source microservice architectures to validate our approach.

Detecting and Resolving Coupling-Related Infrastructure as Code Based Architecture Smells in Microservice Deployments

“…The most recommended option is the (ii) Single Service per Execution Environment pattern [4], in which each service is deployed in its execution environment and can be managed independently. In our previous work [8], we empirically identified two metrics that can be used to differentiate and assess the decision options' conformance:…”

“…This study aims to provide actionable solutions to fix architectural smells of loose coupling-related IaC best practices. We focus on two major Architectural Design Decisions (ADD) in this scope, System Coupling through Deployment Strategy and System Coupling through Infrastructure Stack Grouping, that have been modeled based on an empirical study of existing best practices and patterns used by practitioners in our previous work [8].…”

“…We provide automated architecture refactoring tailored for architectural design in the context of IaC-related ADDs. We also employ the experimentally proven metrics suggested in our previous work [8]. These metrics allow us to analyze the degree to which an IaC deployment model adheres to preferred or less preferred design alternatives for each of the ADDs previously mentioned.…”

“…We implemented automated smell detection and refactoring algorithms to detect potential smells and develop every solution that may be used to solve each smell. The improvements over the initial version are then measured using our metrics [8] on coupling aspects in IaC-based deployments. The results show that every smell can be addressed in no more than three refactoring steps, producing ideal metric values.…”

“…This section will briefly discuss two coupling-related ADDs and their associated options. This information is based on our previous research [8], in which we conducted an empirical study to identify the IaC-related best practices and patterns currently used by practitioners. We also examined the potential decision drivers, or the factors that influence the decisionmaking process, and developed metrics to evaluate how well a given system model adheres to our decision model's recommended patterns and practices.…”