In order to help manage the complexity, heterogeneity and dynamic inherent in distributed real-time systems, middleware of next generation should deal with changing environments and different client requirements. it i...In order to help manage the complexity, heterogeneity and dynamic inherent in distributed real-time systems, middleware of next generation should deal with changing environments and different client requirements. it is very necessary to resolve interoperability between heterogeneous schedulers. Therefore, this promotes the research of a heterogeneous distributed real-time scheduling system based on reflective middleware. The features of reflection technology and reflective middleware are described in detail first. As a result, proposed reflective middleware is used to resolve interoperability between heterogeneous schedulers in distributed real time system. The high performance,adaptability and feasibility of this middleware platform used to resolve interoperability between heterogeneous schedulers are proved in details by a concrete example. Future trends of research in this field are listed at the end.展开更多
为研究异构多核片上系统(multi-processor system on chip,MPSoC)在密集并行计算任务中的潜力,文章设计并实现了一种适用于粗粒度数据特征、面向任务级并行应用的异构多核系统动态调度协处理器,采用了片上缓存、任务输出的多级写回管理...为研究异构多核片上系统(multi-processor system on chip,MPSoC)在密集并行计算任务中的潜力,文章设计并实现了一种适用于粗粒度数据特征、面向任务级并行应用的异构多核系统动态调度协处理器,采用了片上缓存、任务输出的多级写回管理、任务自动映射、通讯任务乱序执行等机制。实验结果表明,该动态调度协处理器不仅能够实现任务级乱序执行等基本设计目标,还具有极低的调度开销,相较于基于动态记分牌算法的调度器,运行多个子孔径距离压缩算法的时间降低达17.13%。研究结果证明文章设计的动态调度协处理器能够有效优化目标场景下的任务调度效果。展开更多
针对异构系统中基于多副本机制的容错调度方法忽略调度makespan、任务间依赖与系统链路失效及严格调度方式调度makespan较长问题,首先提出通用调度方式下同时考虑节点和链路失效的可靠性计算方法;然后给出该通用调度问题的0-1整数规划模...针对异构系统中基于多副本机制的容错调度方法忽略调度makespan、任务间依赖与系统链路失效及严格调度方式调度makespan较长问题,首先提出通用调度方式下同时考虑节点和链路失效的可靠性计算方法;然后给出该通用调度问题的0-1整数规划模型;接着提出可靠性意识多副本任务通用调度(RAMD_TGS,reliability-aware multi-duplication task general scheduling)算法,通过遗传算法种群进化来搜索副本映射节点和开始执行时间。实验表明该算法不仅满足可靠性要求,而且与严格调度方式相比能进一步减小调度makespan,该算法资源占用开销也是可接受的。展开更多
文摘In order to help manage the complexity, heterogeneity and dynamic inherent in distributed real-time systems, middleware of next generation should deal with changing environments and different client requirements. it is very necessary to resolve interoperability between heterogeneous schedulers. Therefore, this promotes the research of a heterogeneous distributed real-time scheduling system based on reflective middleware. The features of reflection technology and reflective middleware are described in detail first. As a result, proposed reflective middleware is used to resolve interoperability between heterogeneous schedulers in distributed real time system. The high performance,adaptability and feasibility of this middleware platform used to resolve interoperability between heterogeneous schedulers are proved in details by a concrete example. Future trends of research in this field are listed at the end.
文摘为研究异构多核片上系统(multi-processor system on chip,MPSoC)在密集并行计算任务中的潜力,文章设计并实现了一种适用于粗粒度数据特征、面向任务级并行应用的异构多核系统动态调度协处理器,采用了片上缓存、任务输出的多级写回管理、任务自动映射、通讯任务乱序执行等机制。实验结果表明,该动态调度协处理器不仅能够实现任务级乱序执行等基本设计目标,还具有极低的调度开销,相较于基于动态记分牌算法的调度器,运行多个子孔径距离压缩算法的时间降低达17.13%。研究结果证明文章设计的动态调度协处理器能够有效优化目标场景下的任务调度效果。
文摘针对异构系统中基于多副本机制的容错调度方法忽略调度makespan、任务间依赖与系统链路失效及严格调度方式调度makespan较长问题,首先提出通用调度方式下同时考虑节点和链路失效的可靠性计算方法;然后给出该通用调度问题的0-1整数规划模型;接着提出可靠性意识多副本任务通用调度(RAMD_TGS,reliability-aware multi-duplication task general scheduling)算法,通过遗传算法种群进化来搜索副本映射节点和开始执行时间。实验表明该算法不仅满足可靠性要求,而且与严格调度方式相比能进一步减小调度makespan,该算法资源占用开销也是可接受的。
