Research On Reliability Task Scheduling Strategy Of Heterogeneous Distributed System

With the rapid development of computer and network technology,distributed systems are more and more widely used in flight control,autonomous vehicle,seismic data processing and other key areas.Heterogeneous distributed systems have great prospects in improving system computing performance and high reliability.Efficient and reliable task scheduling strategy is the key to achieve high performance in heterogeneous distributed systems.This paper mainly focuses on the problem of reliable task scheduling strategy in heterogeneous distributed systems,the main achievements are as follows:First,a new high reliable primary/backup scheduling algorithm called HRPB is proposed to solve the issue when scheduling backup copies with fewer processors where tasks have precedence constraints in heterogeneous distributed systems.The tasks can be modelled as a directed acyclic graph(DAG)and the proposed algorithm schedules two copies of each task.When sorting tasks,it adopts a new index called ALST based on tasks' execution time and deadline;When selecting processor,it considers preparing one more backup copy to solve the issue when scheduling backup copies with fewer processors where tasks have precedence constraints.Second,an improved primary/backup scheduling strategy based on simulate anneal algorithm called SAPB is proposed for task scheduling where tasks have precedence constraints in heterogeneous distributed systems.When sorting tasks,it adopts the method of HEFT to avoid the limitation of the task model such as eFRD;When selecting processor,the simulate anneal algorithm is used to search the solution with higher reliability under the condition of meeting the deadline.Third,a scheduling method considering reliability called MRHS is proposed for hybrid tasks in heterogeneous distributed systems.Hybrid tasks are divided into periodic tasks and aperiodic tasks.Periodic tasks can be modelled as a directed acyclic graph(DAG)and aperiodic tasks are represented by quadruple models.Only periodic tasks are scheduled in the off-line stage,and simulate anneal algorithm is used to search the scheduling results that satisfy the higher defined index value.Randomly entered aperiodic tasks are scheduled in the on-line stage,the processors' idle time slots are used to select the scheduling results with higher defined index value,and the delayed strategy and task reassigned strategy are introduced to solve the issue that all processors' idle time slots do not meet the demand of randomly entered aperiodic tasks' scheduling.Finally,for the reliable task scheduling strategy proposed above,the performance of the proposed reliable task scheduling strategy is verified by simulation tests using randomly generated task model.