Trace partitioning and local monitoring for asynchronous components

Abstract

We propose an instrumentation technique for monitoring asynchronous component systems that departs from the traditional runtime verification set-up assuming a single execution trace. The technique generates partitioned traces that better reflect the interleaved execution of the asynchronous components under scrutiny, and lends itself well to local monitoring. We provide argumentation for the qualitative benefits of our approach, demonstrate its implementability for actor-based systems, and justify claims related to the applicability and efficiency gains via an empirical evaluation over a third party component-based system.