Towards a General Framework for Ensuring and Reusing Proofs of Termination Detection in Distributed Computing

Abstract

Distributed algorithms are designed to run on interconnected autonomous computing entities for achieving a common task: each entity executes asynchronously the same code and interacts locally with its immediate neighbours. It is widely agreed that the lack of knowledge of the global state makes termination detection one of the most important and complex problems in distributed computing. By relying on refinement, we prove that an algorithm computing a spanning tree with Local Termination Detection (each entity is able to determine only its own termination condition), can be reused and adapted in order to compute the same algorithm with Global Termination Detection (at least one entity is aware that the entire computation is achieved in the network). The main idea relies upon specifying a combination of a well known algorithm namely SSP and the spanning tree algorithm, following a top/down approach. This paper is a starting point towards a general framework for enhancing termination detection property of distributed algorithms and reusing their proofs.