Visualising Software Dynamics through Architecture Mining
Summary
When reconstructing software architecture based on data extracted from a running system, the amount of extracted data can become overwhelming (i.e. dozens of gigabytes) and difficult to comprehend. The field of visual analytics builds better and more effective ways to understand and analyse such large datasets. Despite this, there is a lack of research that studies how software execution data can be used to visualise architectural information effectively. Moreover, the gap between research and practice in the field of software architecture visualisation is emphasised by several researchers. We present Architecture Miner, a web-based interactive dashboard that makes it possible to extract architectural intelligence from large-scale Java applications. It provides an overview of design-time structure in relation to run-time processes and metrics, given a scenario chosen by the user. We applied Architecture Miner to the acceptance environment of a 'real-life' Java application with 225,000 LOC, 1,921 classes and 348 packages. Based on an extensive validation with the DevOps team of the system, we can conclude that, within the context of the case study, the designed visualisations are effective in providing valuable architectural information. We consider our results to be a good indication that the insights extracted with Architecture Miner can be valuable within multiple use cases in different contexts.