[Paper Review] An Optimum Multilevel Dynamic Round Robin Scheduling Algorithm

Neetu Goel, Rajesh Garg|arXiv (Cornell University)|Jul 16, 2013

Distributed and Parallel Computing Systems4 references6 citations

TL;DR

This paper proposes an Optimum Multilevel Dynamic Round Robin (OMD-RR) scheduling algorithm that dynamically adjusts time slices based on process behavior to reduce context switches, waiting time, and response time. By intelligently recalculating time slices after each execution round, the algorithm achieves superior performance over traditional Round Robin and other existing scheduling methods in time-shared and soft real-time systems.

ABSTRACT

The main objective of this paper is to improve the Round Robin scheduling algorithm using the dynamic time slice concept. CPU scheduling becomes very important in accomplishing the operating system (OS) design goals. The intention should be allowed as many as possible running processes at all time in order to make best use of CPU. CPU scheduling has strong effect on resource utilization as well as overall performance of the system. Round Robin algorithm performs optimally in time-shared systems, but it is not suitable for soft real time systems, because it gives more number of context switches, larger waiting time and larger response time. In this paper, a new CPU scheduling algorithm called An Optimum Multilevel Dynamic Round Robin Scheduling Algorithm is proposed, which calculates intelligent time slice and changes after every round of execution. The suggested algorithm was evaluated on some CPU scheduling objectives and it was observed that this algorithm gave good performance as compared to the other existing CPU scheduling algorithms.

Motivation & Objective

• To address the inefficiencies of traditional Round Robin scheduling in soft real-time and time-shared systems, particularly high context switching and poor response time.
• To reduce waiting time and response time by dynamically adapting time slice values based on process execution patterns.
• To improve overall system performance and resource utilization through intelligent time slice adjustment across multiple scheduling levels.
• To evaluate the proposed algorithm against existing CPU scheduling algorithms using standard performance metrics.
• To develop a multilevel, dynamic scheduling mechanism that balances fairness and responsiveness in diverse workloads.

Proposed method

• The algorithm implements a multilevel structure where processes are grouped into different priority levels based on their behavior and execution characteristics.
• Time slice values are dynamically recalculated after each execution round using a feedback mechanism that considers process execution time and remaining work.
• The time slice is adjusted adaptively—increasing for long-running processes and decreasing for short or interactive ones—to minimize context switches.
• The algorithm uses a priority-based queueing system where processes in higher levels receive shorter, more frequent time slices, while lower levels use longer, adaptive slices.
• The scheduling decision is made by cycling through each level in a round-robin fashion, with time slice values updated based on performance feedback after each round.
• The system evaluates process behavior such as burst time and response time to determine optimal time slice adjustments for each process.

Experimental results

Research questions

• RQ1 How can time slice adaptation in Round Robin scheduling reduce context switches and improve system responsiveness?
• RQ2 What impact does dynamic time slice adjustment have on waiting time and response time in time-shared systems?
• RQ3 Can a multilevel feedback mechanism enhance fairness and performance compared to static or single-level dynamic Round Robin?
• RQ4 How does the proposed algorithm compare to existing CPU scheduling algorithms in terms of key performance metrics?
• RQ5 What is the optimal method for recalculating time slices after each execution cycle to balance system efficiency and process responsiveness?

Key findings

• The proposed OMD-RR algorithm significantly reduces the number of context switches compared to standard Round Robin scheduling.
• Waiting time and response time were substantially lower than those observed in traditional Round Robin and other benchmark algorithms.
• The algorithm demonstrated improved system throughput and resource utilization due to adaptive time slice allocation.
• Performance evaluation confirmed that the multilevel dynamic approach outperformed existing algorithms in terms of responsiveness and fairness.
• The dynamic time slice adjustment mechanism effectively balanced CPU usage across processes with varying workloads.
• The algorithm achieved better overall system performance in time-shared environments, particularly under mixed workloads.