Task-aware Priority Scheduling For Multicore Processors

With the development of computer software and hardware,cluster system based on multicore processors and parallel processing technology become the focus of the various walks of life.In the course of evolution on parallel processing technology,task scheduling strategy has always been the key issue on multicore processors.For nodes in a parallel framework,task scheduling can be split into task scheduling among nodes and task scheduling within a node.These task scheduling strategies,at present,address primarily the task scheduling problem among nodes.Few studies have focused on the task scheduling analysis within node in parallel framework.Based on the task scheduling technology on multicore processors and considering the particularity of intra-node scheduling in parallel framework,this paper proposes a task-aware priority scheduling for multicore processors,which aims to develop reasonable intra-node scheduling strategy and improve the overall performance of the parallel framework.At the same time,several key issues to be considered are listed as follows: task type,task execution order,task resource allocation and task synchronization.In order to solve these problems,this paper designs and implements the task model which is sensible to task type and task order and the main framework of task-aware priority scheduling,and introduce the scheduling process of this framework four modules: kernel scheduler,resource management,synchronization,node coordinator,together with two key algorithms: task-to-core assignment algorithm and resource allocation algorithm.This main framework mainly completes the related scheduling works such as recognizing the task type,setting the task priority,maintaining the task priority queue,and allocating task resource.Finally,this paper evaluates the performance of task-aware priority scheduling through two groups of labs.Based on the parallel shortest path algorithm,the system performance of two core algorithms under different parameters is designed and implemented.Through comparison with the Work-stealing algorithm,it can be seen that the task-aware priority scheduling strategy achieves a good balance between demand and performance,with the acceptable system performance load.