Observability Design in Microservice Systems

Abstract

Microservice architecture is utilized broadly in modern software development due to its benefits, such as modularity, scalability, focus on business capabilities, implementation flexibility, and faster development and update cadence enabled by modern automation and individual service deployments. However, these benefits come at the cost of increased complexity at the system level, which makes forming a holistic picture of the application and its myriad states difficult.

Observability is a concept originating from control system theory, defined as a measure of how well the internal workings of a system can be inferred from its outputs, but when applied to modern microservice systems this concept has evolved further. Complex distributed systems will inevitably end up in unexpected failure states, and having observability into a system is the only way to methodically investigate the root cause.

This thesis reviews ten literature sources, ranging from scientific articles to technological books to 1) determine best practices for designing microservice systems for observability and 2) collect commonly referenced tools for implementing observability in microservice systems.

As a result, twenty-seven best practices were identified and categorized into five groups: data types and shape, tools and instrumentation, alerting, organization and culture, or integration and lifecycle. In addition, sixty-seven observability tools were collected, organized by times referenced to determine their pervasiveness in literature, and augmented by keywords mined from the context they appeared in, to aid the reader in pursuit of specific tools to integrate into observability solution(s).