关键词: 异构实时仿真, 数据分发服务, 无锁算法, 批处理, 冒险指针

Abstract:

FastDDS faces limitations under high-frequency data streams, such as lock contention, performance overhead from frequent context switching, and configuration complexity of multiple nodes under strict real-time constraints, which affect the experimental efficiency. This paper proposed a performance optimization method based on a batch-scalable circular queue (BSCQ). The approach replaced the traditional mutex mechanism with a lock-free algorithm to reduce lock contention and avoid deadlocks, while batch processing improved data locality, cache hit rates, and memory utilization, effectively reducing data transmission delay and improving system throughput.Hazard pointers were introduced to ensure safe memory management during batch processingand eliminate memory leak risks. Experimental results show that the optimized scheme shows significant advantages in small message and high concurrency scenarios, and its throughput and latency metrics are better than the existing baseline scheme. In addition, this method does not rely on the third-party software environment, has strong versatility and portability, and can be combined with hardware optimization or other high-performance protocols to further improve the system's performance.

Key words: heterogeneous real-time simulation? data distribution service(DDS), lock-free algorithm, batch processing, hazard pointer