摘要
针对实时异构系统的任务调度问题,提出了一种异构多处理器系统的混合实时任务调度算法。该算法采用带有非周期服务器的EDF(Earliest Deadline First)算法来调度单处理器上的任务集,可充分利用处理器的计算带宽。采用启发式搜索算法来进行任务的分配,以最大剩余计算带宽为搜索指标,可确保各处理器的负载尽量平衡。同时,对软实时任务引入QoS(Quality of Service)降级机制,可提高任务集的整体调度成功率。最后,对算法进行了仿真实验,仿真结果证明了算法的有效性。
Aiming at the task scheduling of real-time heterogeneous system, an algorithm for hybrid real-time task scheduling of heterogeneous multiprocessor system was proposed. For scheduling tasks on uniprocessor, Earliest Deadline First （EDF） algorithm with aperiodic server was adopted because it can make full use of computing bandwidth of the processor. The heuristic search algorithm was used for task assignment by taking the maximum remained computing bandwidth as a standard for searching when selecting a processor, thus could equilibrate the load of each processor. Further, the method of degrading QoS was introduced for soft real-time tasks because it can improve scheduling success ratio. At last, simulation experiment was implemented and the result proved the validity of the algorithm.
出处
《电光与控制》
北大核心
2011年第12期39-43,共5页
Electronics Optics & Control
基金
国家"八六三"计划基金资助项目(2007AAJ210)
关键词
多处理器系统
任务调度
异构系统
混合实时任务
muhiprocessor system
task scheduling
heterogeneous system
hybrid real-time task
