Instrumentation and Trace Analysis for Ad-Hoc Python Workflows in Cloud Environments

Knowledge of structure is critical to map legacy workflows to environments suitable to run on the cloud. We present a method which characterizes a workflow structure with the execution trace produced by instrumented logging functionality. The method generates the structure of workflows to support their reuse by permitting their transformation into modern execution environments. The method presented in the paper is implemented for Python workflows and demonstrated in the context of several legacy scientific workflows.