Mining microservice design patterns

Abstract

Building microservices based on design patterns is common practice. Due to the scale and dynamic nature of these applications, engineers usually only have an incomplete mental model of the system. We have developed a methodology that identify instances of well known patterns such as caching or fallbacks by analyzing traces of executions. This is in contrast with most prior work that analyzes source code to mine design patterns. Our preliminary results identifying instances of patterns of interest across several different applications is promising and we discuss the different directions we can explore in this space.