Preempt a Job or Not in EDF Scheduling of Uniprocessor Systems
Jinkyu Lee, Kang G. Shin
Abstract
Jinkyu Lee, Kang G. Shin
Abstract
The earliest-deadline-first (EDF) policy has been widely studied for the scheduling of real-time jobs for its effectiveness and simplicity. However, since each preemption incurs an additional delay to the execution of jobs, the effectiveness of EDF is affected greatly by the underlying preemption policy that determines if and when a higher-priority job is allowed to preempt a currently executing lower-priority job. To address this problem, we propose a new and better (in meeting job deadlines) preemption policy of EDF, given a non-zero preemption delay. Specifically, we propose a controlled preemption (CP) policy that controls the condition of preempting jobs, whereas existing approaches focus on that of preempted jobs. We define cp-EDF in which the CP policy is applied to EDF, and analyze its schedulability. This schedulability analysis is then utilized to develop an algorithm that assigns the optimal control parameters of cp-EDF. Our in-depth evaluation has demonstrated that cp-EDF with the optimal parameter assignment improves EDF schedulability over existing preemption policies by up to 7.4%.
OpenAlex reports 18 citations for this work. Citation counts describe recorded attention and do not establish research quality.
A contribution statement is not available in the OpenAlex record.
Method details are not available in the OpenAlex metadata.
Findings are not separately available in the OpenAlex metadata.
Limitations are not available in the OpenAlex metadata.
Application details are not available in the OpenAlex metadata.
The earliest-deadline-first (EDF) policy has been widely studied for the scheduling of real-time jobs for its effectiveness and simplicity. However, since each preemption incurs an additional delay to the execution of jobs, the effectiveness of EDF is affected greatly by the underlying preemption policy that determines if and when a higher-priority job is allowed to preempt a currently executing lower-priority job. To address this problem, we propose a new and better (in meeting job deadlines) preemption policy of EDF, given a non-zero preemption delay. Specifically, we propose a controlled preemption (CP) policy that controls the condition of preempting jobs, whereas existing approaches focus on that of preempted jobs. We define cp-EDF in which the CP policy is applied to EDF, and analyze its schedulability. This schedulability analysis is then utilized to develop an algorithm that assigns the optimal control parameters of cp-EDF. Our in-depth evaluation has demonstrated that cp-EDF with the optimal parameter assignment improves EDF schedulability over existing preemption policies by up to 7.4%.
Key concepts: Preemption, Uniprocessor system, Computer science, Scheduling (production processes), Earliest deadline first scheduling, Execution time, Worst-case execution time, Real-time operating system