Abstract:

Nowadays, the traditional serial programs for TT&C data storage system is becoming difficult to adapt to the high load task demands in the new environment of continuously increased instrumentation equipment’s numbers and data rates. In this paper, a parallel real-time TT&C data storage system framework is designed and implemented based on OpenMP and Qt with the parallel idea of SPMD（Single Program Multiple Data）. By introducing the M/M/1 queuing model, a static task partitioning algorithm is designed to balance the load between parallel threads. At the same time, a fine-grained synchronization mechanism between threads is completed for shared memory communication based on mutex locks. And the signal and slot mechanism of Qt is successfully embedded into OpenMP threads by the means of event loop, thus the good man-computer interaction capability of Qt is retained in the parallel framework. Experimental results show that the parallel framework has better scalability and real-time performance compared with the serial TT&C data storage system under high load, which can make a better use of the computing resources of the multi-core system and obtain better real-time performance.

Key words: Single Program Multiple Data（SPMD）；parallel real-time TT&C data storage system, load balance, mutex locks

摘要：

随着测控设备数量增加、数据码率的不断提高，传统的测控中心实时存储系统采用的串行化处理逻辑已经难以适应高负载的测控任务需要。采用SPMD（Single Program Multiple Data）思想，基于OpenMP和Qt设计并实现了一套并行实时测控数据存储系统框架。该系统通过引入M/M/1模型的排队系统，设计了一种适合并行实时测控数据存储系统的静态负载均衡算法，基于互斥锁完成了线程间的细粒度通信同步机制，借助于事件循环机制将Qt的信号与槽特性成功地嵌入OpenMP线程，保留了Qt图形界面良好的人机交互性能。实验结果表明该系统在高负载的情况下，相比串行测控数据存储系统而言具有更好的扩展性和实时性，能更好地利用现行单机多核系统的计算资源，获得更好的实时性能。

关键词: 单程序多数据（SPMD）, 并行测控数据实时存储系统, 负载均衡, 互斥锁