Industry practices and challenges for the evolvability assurance of microservices

Abstract

Context Microservices as a lightweight and decentralized architectural style with fine-grained services promise several beneficial characteristics for sustainable long-term software evolution. Success stories from early adopters like Netflix, Amazon, or Spotify have demonstrated that it is possible to achieve a high degree of flexibility and evolvability with these systems. However, the described advantageous characteristics offer no concrete guidance and little is known about evolvability assurance processes for microservices in industry as well as challenges in this area. Insights into the current state of practice are a very important prerequisite for relevant research in this field. Objective We therefore wanted to explore how practitioners structure the evolvability assurance processes for microservices, what tools, metrics, and patterns they use, and what challenges they perceive for the evolvability of their systems. Method We first conducted 17 semi-structured interviews and discussed 14 different microservice-based systems and their assurance processes with software professionals from 10 companies. Afterwards, we performed a systematic grey literature review (GLR) and used the created interview coding system to analyze 295 practitioner online resources. Results The combined analysis revealed the importance of finding a sensible balance between decentralization and standardization. Guidelines like architectural principles were seen as valuable to ensure a base consistency for evolvability and specialized test automation was a prevalent theme. Source code quality was the primary target for the usage of tools and metrics for our interview participants, while testing tools and productivity metrics were the focus of our GLR resources. In both studies, practitioners did not mention architectural or service-oriented tools and metrics, even though the most crucial challenges like Service Cutting or Microservices Integration were of an architectural nature. Conclusions Practitioners relied on guidelines, standardization, or patterns like Event-Driven Messaging to partially address some reported evolvability challenges. However, specialized techniques, tools, and metrics are needed to support industry with the continuous evaluation of service granularity and dependencies. Future microservices research in the areas of maintenance, evolution, and technical debt should take our findings and the reported industry sentiments into account.

Static analysis tools in the era of cloud-native systems

“…A large set of challenges was identified by practitioners in a large study on microservice evolvability [1]. Missing system-centric perspectives, inter-service dependencies, coordination between decentralized teams, and challenges with outdated documentation, were the most frequently mentioned issues.…”

Microservices-based systems visualization

“…That requires engineers to understand what happened across the entire service graph at the time of the debugging, which seems to counter to the ethos of microservices architectures in the first place. Bogner et al focused on this challenge in a systematic grey literature review [3]. They concluded that most painful challenges in the microservices industry are the Service-cutting and No-System-Centric views.…”

Microservice Architecture Reconstruction and Visualization Techniques: A Review

“…Microservice development comes from practitioners, and research tends to come later, so publications about microservices are still limited in a lot of areas. Thus, grey literature may hold valuable insights that academic literature simply cannot provide yet [52].…”