Distributed Reactive Programming for Reactive Distributed Systems

Florian Myter¹, Christophe Scholliers², and Wolfgang De Meuter³

The Art, Science, and Engineering of Programming, 2019, Vol. 3, Issue 3, Article 5

Submission date: 2018-06-01
Publication date: 2019-02-01
DOI: https://doi.org/10.22152/programming-journal.org/2019/3/5
Full text: PDF

Abstract

Context: The term reactivity is popular in two areas of research: programming languages and distributed systems. On one hand, reactive programming is a paradigm which provides programmers with the means to declaratively write event-driven applications. On the other hand, reactive distributed systems handle client requests in a timely fashion regardless of load or failures.

Inquiry: Reactive programming languages and frameworks tailored to the implementation of distributed systems have previously been proposed. However, we argue that these approaches are ill fit to implement reactive distributed systems.

Approach: We analyse state of the art runtimes for distributed reactive programming and identify two key issues with regards to reactive distributed systems. They rely on single, central points of coordination and/or assume a lack of partial failures in the systems they support.

Knowledge: Based on our analysis we propose a novel runtime for distributed reactive programming languages and frameworks. This runtime supports reactive distributed systems by design.

Grounding: We implement a proof of concept framework for reactive distributed systems in JavaScript which builds atop our runtime. Using this framework we implement a case study application which highlights the applicability of our approach. Moreover, we benchmark our runtime against a similar approach in order to showcase its runtime properties and we prove its correctness.

Importance: This work aims to bridge the gap between two kinds of reactivity: reactive distributed systems and distributed reactive programming. Current distributed reactive programming approaches do not support reactive distributed systems. Our runtime is the first to bridge this reactivity gap: it allows for reactive distributed systems to be implemented using distributed reactive programming.