Scheduling real-time tasks with reduced context switches.

摘要

Computer systems that manage many time-critical tasks are referred to as real-time systems. Computer systems of this kind require efficient and simple scheduling algorithms. Hence, scheduling real-time tasks has been an important topic in real-time systems for a long time. Real-time scheduling problems may involve either fixed-duration tasks or recurring tasks that must be completed within a certain time frame. The problems most studied within the recurring category involve periodically recurring tasks. Several approaches to scheduling real-time tasks have been investigated.;Even if much progress has been made in regard to preemptive real-time scheduling algorithms, there are relatively few results in the case of scheduling algorithms that reduce context switches that cause an increase in overhead in real-time applications. This observation has motivated our development of efficient scheduling algorithms in terms of reducing context switches in real-time systems.;In this thesis, we present priority-driven and preemptive scheduling algorithms for periodic, bounded predictable, and soft aperiodic real-time tasks with reduced context switches. First, we discuss the problem of scheduling periodic tasks with reduced context switches. Second, we consider the problem of scheduling periodic tasks in soft aperiodic environments. The proposed solutions for joint scheduling of periodic and soft aperiodic tasks provide the same schedulability with significantly decreased context switches compared to other available solutions while providing good response times for aperiodic tasks. Next, we investigate the problem of joint scheduling of periodic and bounded predictable tasks under the principle of reducing context switches. The proposed solutions should guarantee the schedulability of tasks with significantly fewer context switches compared to existing solutions. Finally, we consider the problem of joint scheduling of periodic, aperiodic, and bounded predictable tasks with reduced context switches. The operation and performance of the proposed algorithms are discussed and compared with existing alternative solutions in terms of context switches. A valuable contribution of this research is to provide the real-time system designer with a wide range of alternatives for designing real-time operating systems.