A Fresh Look at Modeling Distributed Reactive Systems

Scalable, distributed reactive systems require a scalable control infrastructure. Being able to model such an infrastructure is a prerequisite for putting it into place. In current systems, finite state machine based approaches are prevailing. They are, however, limited by their nature and hamper further improvements. Extensions like state charts have emerged to alleviate those shortcomings. Despite all this work, the proposed improvements do not live up to the requirements of continuously operating distributed systems. Therefore we look for alternatives that break with the underlying state-based model. Workflow systems recently became mature enough to be considered for use cases going beyond small-scale business applications. The concept on which they build is promising to scale to large systems. This paper compares various modeling technologies using the real-world scenario of distributed data acquisition systems as we find them in currently operating high-energy physics installations at the Large Hadron Collider at CERN. The outcome of our qualitative evaluation will show that current finite state machine compared to workflow based approaches have a bigger gap between specification and executed processes. Consequently workflow based approaches are better suited for the use case at hand and require less custom software extensions to reduce the remaining gap.