Evaluation of CPU Scheduling and Synchronization Algorithms in Distributed Systems

Abstract

This paper presents an evaluation model designed to compare various CPU scheduling and synchronization algorithms in the context of a distributed system. A simulation model is constructed with six nodes, each equipped with two active and three passive resources, and featuring a periodic task per node. Despite limitations on simulation scale due to resource constraints, the study applies four CPU scheduling algorithms and finds that the Earliest Deadline First (EDF) algorithm outperforms others in terms of system performance. Additionally, the paper explores synchronization algorithms, highlighting the CMH algorithm's focus on snapshot collecting, and distinguishing between 2PC and 3PC for transactional operations. Furthermore, the study discusses the characteristics of Raft and Paxos in ensuring consistency during distributed system operations, with Raft emphasizing identical indexes and terms for identical operations, and Paxos favoring the candidate with the higher term in the event of simultaneous candidacy.