船舶仿真计算集群智能调度系统的开发

Development of Intelligent Scheduling System of Ship Simulation Computing Cluster

针对上船院(SDARI)高性能计算平台静态调度方式在研发和设计过程中不能满足当前实际需求等问题,基于FCFS(first come first service)和Firstfit算法提出一种优化的FCFS-Fit调度算法。通过对集群进行数据采集,负载预测分析,为调度系统提供决策依据,并在网页端开发权限管理和交互式可视化功能模块。通过该系统实现了高性能计算平台的智能调度,大大提高了集群的使用效率,提升了高性能调度的精细化管理水平,系统稳定性较好,具有一定的工程实际应用价值。

Aiming at the problem that the static scheduling mode of high-performance computing(HPC) platform of Shanghai merchant ship design and research institute can not meet the current actual needs during the development and design process,based on FCFS(first come first service) algorithm and Firstfit algorithm,an optimized FCFS-Fit scheduling algorithm was proposed.Through the data collection and load prediction analysis of the cluster,a decisionmaking basis was provided for the scheduling system.On the web side,permission management and interactive visualization function modules had been developed.Through this platform,the intelligent scheduling of high-performance computing platforms was realized.The efficiency of cluster usage was greatly improved.And the refined management level of high-performance scheduling was promoted.The system had good stability and certain engineering practical application value.