Towards a Technique for Extracting Relational Actors from Monolithic Applications

Relational actors, or reactors for short, integrate the actor model with the relational data model, providing an abstraction for enabling actor-relational database systems. However, as a novel model of computation for databases, there is no extensive work on reasoning about reactor modeling. To close this gap, this paper aims to review methods for systems decomposition in order to analyze their completeness and propose as well as evaluate a technique for reactor modeling. Concretely, we put forward a technique to extract reactors from a monolithic system. For evaluation, we selected a REST-based open-source OLTP system in which a decomposition to microservices was conducted and applied our technique on its predecessor monolithic version. Our technique led to the same set of decisions, regarding table and behavior selection, taken by experts when decomposing the same system into microservices. The proposed technique can be seen as a first step towards supporting practitioners in decomposing OLTP systems into reactors.