COMPARATIVE ANALYSIS OF MODIFIED FIRST-COME FIRST-SERVED AND REACTIVE MULTI-RESOURCE TOKEN – EARLIEST DEADLINE FIRST ALGORITHMS UNDER LIMITED ACCESS TO MAIN MEMORY

DOI: 10.31673/2412-4338.2025.038704

Abstract

This paper presents a comparative experimental study of the efficiency of two task scheduling algorithms under conditions of limited access to main memory: the Modified First-Come First-Served (MFСFS) algorithm and the hybrid strategy known as Reactive Multi-Resource Token – Earliest Deadline First (RMRT-EDF). As part of the study, an enhancement to the classical FCFS algorithm is proposed, incorporating a pre-execution check of memory availability. This modification reduces the likelihood of system overloads and scheduler failures. However, in cases where the first task in the queue cannot be executed due to insufficient memory—despite the presence of less resource-intensive tasks—the system experiences a blocking effect, which negatively impacts overall performance.

In contrast, the RMRT-EDF algorithm implements an adaptive scheduling policy that takes into account both task deadlines and the current state of system resources. Based on the concept of resource tokens, this approach prioritizes tasks with the earliest deadlines among those that fit within the available resources. Additionally, a resource

monitoring module with a 500 ms update interval is employed, providing real-time memory usage data and enabling the algorithm to dynamically adapt to workload fluctuations.

To evaluate the algorithms' effectiveness, a test environment was developed to simulate high-intensity workloads involving over 10,000 tasks with varying resource demands. The experiments were conducted in two modes: with and without memory constraints. The results demonstrated that under memory-limited conditions, RMRT-EDF significantly outperforms MFСFS in the number of successfully completed tasks without SLA violations, while maintaining a comparable level of overall performance.

The findings confirm the viability of using reactive, adaptive scheduling strategies in resource-constrained environments and lay the groundwork for further research into real-time dispatching optimization.

Keywords: task scheduling, resource constraints, FCFS, RMRT-EDF, deadline, SLA violation, business services.