COMPARATIVE ANALYSIS OF RESOURCE SCHEDULING ALGORITHMS IN CLOUD COMPUTING ENVIRONMENTS

Abstract

Summary. Introduction. The rapid growth of cloud technologies and large-scale digitalisation
of the economy have made efficient management of computing resources one of the key factors in the
competitiveness of modern IT infrastructures. The problem of effective resource allocation directly
affects compliance with Service Level Agreements (SLA) and the operational costs of cloud service
providers. Purpose of this article is to study the effectiveness of classical and heuristic task scheduling
algorithms in cloud environments based on simulation modelling and to develop a comprehensive
methodology for their quantitative evaluation.
Results. A modular Python-based software complex was developed to conduct simulation
experiments at three load levels: Low Load (10 tasks), Medium Load (20 tasks), and High Load (40
tasks) across clusters of 2 to 6 servers. Five algorithms were compared: FCFS, Round Robin, LPT,
SPT, and Weighted Round Robin. A composite efficiency metric, Score, was proposed to jointly
evaluate schedule quality (Makespan) and computational overhead of the scheduling algorithm itself.
Experimental results confirm that the LPT algorithm achieves the best Makespan values under
medium and high load conditions, while WRR delivers the best integral Score in heterogeneous server
configurations. No universal optimal algorithm was identified.
Conclusion. The choice of scheduling algorithm must depend on system priorities: LPT is
recommended for batch processing workloads where minimising total execution time is critical; WRR
– for heterogeneous infrastructures where balanced resource utilisation is paramount; Round Robin
and FCFS – for lightweight real-time scenarios with strict latency constraints. The proposed Score
metric provides a practical tool for adaptive algorithm selection in Cloud Management Systems.