基于预留的信息物理融合系统动态内存分配方法研究

Research on a Dynamic CPS Memory Allocation Method Based on Reservation

信息物理融合系统(CPS)是一种深度嵌入式分布式实时系统,时效性是其关键属性。由于CPS的节点中内存资源有限,当多个实时任务并发执行竞争内存时,将导致任务错过截止期,严重影响CPS的时效性。针对该问题,提出了基于预留的动态内存分配方法。通过为并发执行的实时任务分配私有预留内存块和共享预留的内存块的方法,对内存进行高效管理,有效避免了任务竞争内存资源导致的系统时效性下降的问题。实验结果表明,在有限的内存资源环境中,提出的CPS动态内存分配方法有着更低的截止期错失率以及更高的系统稳定性。

Cyber and physical system is a deeply embedded distributed real-time system whose the key attribution is timeliness. Because of the limited resource on CPS nodes, when parallel real time tasks compete the memory, the task will miss the deadline which affects the timeliness of CPS badly. Aiming at this problem, this paper proposed a dynamic memory allocation method based on reservation. The algorithm allocates the private and shared memory block for real time tasks in order to manage memory efficiently, which can effectively avoid the low system performance because of the memory competition among tasks. The result shows that the CPS dynamic memory allocation method maintains the lower miss ratio and higher stability on condition of limited memory resources.