Research On Periodic Task Scheduling Technology

The communications industry has experienced rapid development during these years.Future communications applications will become more and more complex,the number of tasks will increase,and communication systems will increasingly use multi-core system-on-chip to improve computing capabilities.At the same time,with the rapid development of the microelectronics industry,the number of cores integrated in a multi-core system-on-chip is increasing,and the single-core processing capability is also increasing.However,these advancements all increase the difficulty of designing the scheduling algorithm.The task scheduling results will significantly affect the multi-task execution efficiency of multi-core systems in communication systems.A well-performing scheduling algorithm has become the key to system design.The baseband processing tasks in the communication system,the data acquisition tasks in the radar system,and the computation tasks in the integrated modular avionics system are strictly periodic tasks.In order to make full use of the parallel processing capability of multi-core system-on-chip to solve the scheduling problem of non-preemptive strictly periodic tasks in multi-core systems,this paper studies the periodic task scheduling technology.Periodic task scheduling can be mainly divided into two categories: unrelated periodic task scheduling and associated periodic task scheduling.Unrelated periodic tasks refer to periodic tasks with no dependencies between tasks.Associated periodic tasks refer to periodic tasks with dependencies between tasks.These two categories are for single-core high-performance homogeneous multi-core platform based on the x86 platform and single-core relatively weak performance heterogeneous multi-core platform based on the ARM platform respectively.The paper first introduces the previous research that researchers have done in the direction of periodic task scheduling.Combined with the common multi-core platform architecture,the schedulability analysis conditions for periodic tasks are given.Three new scheduling algorithms which are based on this schedulability analysis are proposed.1)MILP algorithm which is a direct-solved scheduling algorithm for unrelated periodic task models2)STA algorithm which is a search-based scheduling algorithm for unrelated periodic task models3)PILP algorithm which is a direct-solved scheduling algorithm for associated periodic task modelsAmong them,the direct-solved scheduling algorithm has low efficiency and the optimal solution.The search-based scheduling algorithm has higher execution efficiency and cannot obtain the optimal solution.The two algorithms are applicable to scenarios with high requirements for scheduling results and calculation time respectively.The choice of the linear programming solver and the optimization of the linear programming model will affect the efficiency.The traditional evaluation indicator of the scheduling algorithm is the schedule length.However,the evaluation based on this indicator is not comprehensive enough.Therefore,this paper also introduces the power consumption,number of active cores,acceptance ratio and traffic as evaluation indicators of the scheduling algorithms.Each evaluation indicator were separately described according to different target platform architecture.According to the evaluation model introduced in this paper,the proposed algorithm and its comparison algorithm are simulated and evaluated.It shows that the algorithm proposed in this paper has advantages in many aspects,and the feasibility and effectiveness of the proposed algorithm are verified.