Agent-Based Adaptive Dynamic Round Robin (AADRR) Scheduling Algorithm

Scheduling techniques are essential to increase resource utilization and task execution within modern computing environments. Round Robin Scheduling (RR) ensures a fair distribution of processes needing attention but often leads to inefficiencies in systems with heterogeneous tasks or different priorities due to large latency or resource usage differences. To address such problems, this paper introduces the Agent-based Adaptive Dynamic Round Robin (AADRR) process scheduling technique, which enhances process scheduling by continuously adjusting the time quantum and criteria, combining CPU burst time and priority. In the proposed AADRR, all processes are ranked dynamically by a software agent based on user preferences and current system load. This agent operates independently by keeping an eye on system parameters and making the required adjustments in real-time without requiring human intervention. We place processes in the queue according to their order of importance. A dynamic time quantum policy is suitable whenever it meets the mean duration of each process in the queue. Every round has the time quantum adjusted based on this method average burst time. AADRR highlights that the short processes are managed properly and the long processes are completed within a few rounds to fairly complete and maintain all the processes in the system. The proposed AADRR is more suitable for periodic tasks that employ a dynamic scheduling system and adapt time quantum according to the system state and job features. Additionally, AADRR efficiently manages preemptable tasks, using dynamic scheduling policies to accommodate variations in process priority and CPU burst times, ensuring fair scheduling, efficient resource utilization, and dynamic adaptability. To validate the effectiveness of the AADRR algorithm, we performed a comparative performance analysis with twelve other algorithms, including five traditional CPU scheduling algorithms and seven advanced job scheduling techniques, demonstrating optimal performance results. In our experiments, synthetic workload traces were generated using the Monte Carlo probability distribution method, which is scientifically recommended for creating diverse workload traces. Small, medium, and large datasets were used, with the small workload traces obtained from published studies and the large traces produced by the Monte Carlo simulation. AADRR efficiently reduces average turnaround times and average waiting times for each workload and performs better in response time. AADRR may not always provide the most favorable measures in all scenarios. Still, it performs better than other scheduling techniques in system performance, while being more efficient and flexible for different workloads.