Simulation And Performance Analysis Of GEDF Scheduling In Parallel Soft Real-time System

With the development of the technology,real-time systems are increasingly used in various emerging fields and applications,such as some trajectory tracking systems and virtual reality systems,and these new application scenarios also put forward some new indicators and requirements for real-time systems.Such real-time systems are called soft real-time systems,and with the growing popularity of multi-core systems,real-time scheduling of multi-core parallel tasks becomes more and more important.However,since the prior work mainly focuses on the non-parallel task model,it poses a new challenge.It is precisely because of the important role displayed by real-time systems of parallel tasks that academic research pays more and more attention to them.Among them,real-time system scheduling algorithm is a research hotspot,and this paper is aimed at simulation and performance analysis of GEDF in parallel soft real-time system.Prior work on GEDF mainly focused on the hard real-time and non-parallel task systems.To ensure that all time requirements were met,in the worst case,approximately 50%of the available processing capacity must be sacrificed.This may be overkill for soft real-time systems that can tolerate bounded tardiness.In this paper,when the task utilization in the parallel task system satisfies u_i?m/2 and the critical path length of the task satisfies L_i<d₁/2,a bounded tardiness is obtained under the preemptive GEDF on the multiprocessor.The bounded tardiness depends on the utilization and execution time of each task which means the lower these values,the lower the tardiness bound.At the same time,this paper also implements a highly scalable DAG-based parallel task scheduling simulator,and implements GEDF,FS,and GEDF_modify scheduling algorithms on it.Finally,the influencing factors affecting the performance of parallel task GEDF scheduling are verified by simulation experiments.It is also proved that GEDF scheduling is more suitable for multi-core soft real-time parallel task system than FS scheduling.