摘要
为了提高HLA仿真系统在高性能计算平台上的运行性能,针对高性能计算平台的高性能层次式通信环境,提出了一种共享内存/IB/以太网自适应RTI通信机制———SACM机制,该机制能够根据通信双方所在节点的网络拓扑信息,自适应建立延迟最小的通信连接,实现共享内存、IB等高速互连及以太网互连的自适应选择。同时在该通信机制中,针对节点内高速共享内存通信需求,提出了一种面向多盟员互连的共享内存通信算法,有效提高了节点内盟员通信效率。对象属性值更新延迟测试结果表明,该通信机制可以有效提高RTI在高性能计算平台的通信性能。
To promote the execution efficiency of HLA system which executes on high performance computing （HPC） platform, a self-adaptive communication mechanism, named SACM, is proposed for the hierarchy communication environment of HPC. By this mechanism, the connection with lowest latency can be established from the network information of computing nodes. The shared memory, high speed IB network and Ethemet network can be selected by federates adaptively. Aiming at the demand of shared memory communication between federates within one node, a high speed shared memory communication algorithm based on distributed infrastructure is proposed, and the communication performance can be improved by this algorithm. Experiment result of object attributes value updating latency has shown that the communication performance of RTI executing on HPC platform has been greatly improved.
出处
《国防科技大学学报》
EI
CAS
CSCD
北大核心
2012年第3期148-153,共6页
Journal of National University of Defense Technology
基金
国家自然科学基金资助项目(61170047)
关键词
分布式交互仿真
运行支撑环境(RTI)
高性能计算平台
自适应通信机制
共享内存通信
distributed interactive simulation
runtime infrastructure
high performance computing platform
self-adaptive communication mechanism
shared memory communication
