面向路径搜索的多虚拟机协同计算关键技术

摘要

研究了一种面向路径搜索的多虚拟机协同计算系统,为计算机辅助工程领域路径搜索问题的求解提供了一种灵活高效的计算环境。针对任务规模和搜索路径的不确定性等,构建了虚拟化平台下的自适应性体系结构和协同工作流程,提出了所需关键技术。针对路径搜索引擎的特殊需求,设计了启发式探测性任务分解算法,以充分挖掘任务在数据、功能等方面的并行性。通过调度多个子任务在多台虚拟机上并行执行,最大限度地提高资源的利用率,实现以资源换取较低的运行时间,加快得到最优解或满意解。设计了启发式解空间优化算法,以优化计算过程和结果数据的管理。通过监控任务的状态,可以避免资源浪费和重复路径执行,进一步减少运行时间。开发了系统引擎模块原型,并以一个计算机辅助工程领域的板材切割路径搜索问题对所设计的算法进行了实验评估。进一步将多虚拟机协同计算系统与四种系统进行对比分析,实验结果表明,多虚拟机协同计算系统比其他系统更能提供平稳的运行状态,并具有较高的性能和效率,原型很好地实现了面向路径搜索任务的求解环境。%To provide a flexible and effective computing environment for solving the path search problems in Computer Aided Engineering（CAE）,a Collaborative Computing System over Multiple Virtual Machines（CCSMVM） was studied.Based on the uncertainty of the task scale and the search path,a self-adaptive architecture and collaborative workflow in virtualized platform were constructed,and the key technologies were also proposed.Aiming at the specific requirements of path search engine,a heuristic algorithm for exploratory tasks decomposition to mine the parallelism in data and functions was designed.By scheduling parallel execution of multiple subtasks on multiple virtual machines,utilization rate of resources was improved greatest so as to exchange resources for a shorter execution time and the optimal solutions or satisfactory solutions were obtained quickly.A heuristic algorithm for solution space optimization was designed to optimize the computing process and result data management.By monitoring status of the tasks,the waste of resources and the duplication of execution paths were avoided,therefore the execution time was further reduced.A prototype system with path search engine module was developed.A path search problem of sheet cutting in CAE field was taken as an experiment appraisal example to verify the algorithms,and compared the performances of CCSMVM with four other systems further.The experimental results showed that the CCSMVM could provide stable execution status with high performance and efficiency,and the prototype realized good solution environment for problems oriented to path search tasks.