Determinism of Replicated Distributed Systems–A Timing Analysis of the Data Passing Process
Adriano A. Santos, António Ferreira da Silva, António P. Magalhães, Mário de Sousa · Advances in Science Technology and Engineering Systems Journal · 2020
Fault-tolerant applications are created by replicating the software or hardware component in a distributed system.Communications are normally carried out over an Ethernet network to interact with the distributed/replicated system, ensuring atomic multicast properties.However, there are situations in which it is not possible to guarantee that the replicas process the same data set in the same order.This occurrence will lead to inconsistency in the data set produced by the replicas, that is, the determinism of the applications is not guaranteed.To avoid these inconsistencies, a set of Function Blocks has been proposed which, taking advantage of the inherent properties of Ethernet, can guarantee the synchronism and determinism of the real-time application.This paper presents this set of Function Blocks, focusing our action on the development of reliable distributed systems in real-time.This demonstrates that the developed Function Blocks can guarantee the determinism of the replicas and, as such, that the messages sent are processed, in the same order and according to the time in which they were made available.