1283-GRP Adaptive Optimized Round Robin: Evaluating Time Quantum Prediction in an Integrated Computer-System Simulation Model

Presenter Information

Location

https://campus.kennesaw.edu/colleges-departments/ccse/events/computing-showcase/fa26-cday-program.php

Document Type

Event

Start Date

20-11-2026 2:00 PM

Description

The performance of Round Robin CPU scheduling depends on the time quantum. The Optimized Round Robin algorithm predicts the time quantum with multiple linear regression. But it was evaluated on static process sets with known CPU burst times and no context switch cost. This paper proposes Adaptive Optimized Round Robin, which estimates CPU burst times with exponential averaging, re-computes the time quantum at every dispatch, and trains its regression model with the context switch cost included. Adaptive Optimized Round Robin is evaluated in a discrete-event simulation model of a complete computer system with random arrivals, memory admission, CPU scheduling, disk I/O, and context switch cost by using 720 simulations with ten paired workload seeds per configuration. At a CPU load of 0.8 and a context switch cost of 0.5 ms, static Optimized Round Robin had a 65–80 ms higher average waiting time than Round Robin with a 10, 20, or 40 ms time quantum. Adaptive Optimized Round Robin had a lower average waiting time than Round Robin with a 5, 10, or 40 ms time quantum and was not significantly different from Round Robin with a 20 ms time quantum. At a context switch cost of 2 ms, Adaptive Optimized Round Robin still had a lower average waiting time than the 5, 10, and 20 ms time quanta, but a 40 ms time quantum gave a 100 ms lower average waiting time. Therefore, the Adaptive Optimized Round Robin avoids the worst fixed time quanta without matching the best one at every setting.

This document is currently not available here.

Share

COinS
 
Nov 20th, 2:00 PM

1283-GRP Adaptive Optimized Round Robin: Evaluating Time Quantum Prediction in an Integrated Computer-System Simulation Model

https://campus.kennesaw.edu/colleges-departments/ccse/events/computing-showcase/fa26-cday-program.php

The performance of Round Robin CPU scheduling depends on the time quantum. The Optimized Round Robin algorithm predicts the time quantum with multiple linear regression. But it was evaluated on static process sets with known CPU burst times and no context switch cost. This paper proposes Adaptive Optimized Round Robin, which estimates CPU burst times with exponential averaging, re-computes the time quantum at every dispatch, and trains its regression model with the context switch cost included. Adaptive Optimized Round Robin is evaluated in a discrete-event simulation model of a complete computer system with random arrivals, memory admission, CPU scheduling, disk I/O, and context switch cost by using 720 simulations with ten paired workload seeds per configuration. At a CPU load of 0.8 and a context switch cost of 0.5 ms, static Optimized Round Robin had a 65–80 ms higher average waiting time than Round Robin with a 10, 20, or 40 ms time quantum. Adaptive Optimized Round Robin had a lower average waiting time than Round Robin with a 5, 10, or 40 ms time quantum and was not significantly different from Round Robin with a 20 ms time quantum. At a context switch cost of 2 ms, Adaptive Optimized Round Robin still had a lower average waiting time than the 5, 10, and 20 ms time quanta, but a 40 ms time quantum gave a 100 ms lower average waiting time. Therefore, the Adaptive Optimized Round Robin avoids the worst fixed time quanta without matching the best one at every setting.