集群计算在数字流域模型中的应用
|
摘要
随着计算机技术的发展,以消息传递接口(MPI)标准为代表的高性价比集群计算技术使并行计算在大量传统的专业领域也得到了广泛应用。数字流域模型因需要对大范围流域进行分布式的水文、泥沙过程模拟而提出了较大规模的计算需求。同时,基于分水岭的单元划分方式和专门的河网编码方法使数字流域模型的并行化计算具有先天优势。提出了一种典型的并行调度流程,用于完成产汇流计算的动态任务分配。在自主搭建的MPI计算集群上进行的应用实验表明,集群计算提高了数字流域模型的计算效率,能够作为模型的计算平台。最后指出了此应用计算平台的发展方向。
With the development of the computing technology,the high performance/price ratio parallel computing,represented by the Message Passing Interface(MPI) standard,is applied widely to the more traditional professional fields.The massive computing is required since that the digital watershed model discussed in this paper is aimed to simulate the distributed hydrological and sediment process in a large-scale watershed.The parallel computing of the digital watershed model has its proper advantage including the cell partition mode based on the watershed division and the technical coding method of the drainage network.This paper presents a typical parallel scheduling flow to accomplish dynamically task assignment in rainfall-runoff and discharge routing computing.The application experiment on the MPI computing cluster constructed in this paper indicates that the cluster computing improves the computing efficiency of the digital watershed model.The paper presents the future development of this cluster computing flat as well.
With the development of the computing technology,the high performance/price ratio parallel computing,represented by the Message Passing Interface(MPI) standard,is applied widely to the more traditional professional fields.The massive computing is required since that the digital watershed model discussed in this paper is aimed to simulate the distributed hydrological and sediment process in a large-scale watershed.The parallel computing of the digital watershed model has its proper advantage including the cell partition mode based on the watershed division and the technical coding method of the drainage network.This paper presents a typical parallel scheduling flow to accomplish dynamically task assignment in rainfall-runoff and discharge routing computing.The application experiment on the MPI computing cluster constructed in this paper indicates that the cluster computing improves the computing efficiency of the digital watershed model.The paper presents the future development of this cluster computing flat as well.
引文
[1]陈国良.并行计算-结构、算法、编程[M].北京:高等教育出版社,1999.
[2]Top500 List for June 2005[EB/OL],http://www.top500.org/lists/2005/06/
[3]田华,卞建春,颜宏.浅水波模式半拉格朗日方法的并行研究[J].应用气象学报,2004,15(4):417-426.
[4]袁金南,王在志,薛纪善.广州区域数值预报模式并行化计算[J].应用气象学报,2004,15(5):556-563.
[5]赵改善,包红林.集群计算技术及其在石油工业中的应用[J].石油物探,2001,40(3):118-126.
[6]赵改善,孔祥宁,王于静,等.64位集群计算平台波动方程叠前深度偏移的性能优化[J].勘探地球物理进展,2005,28(1):57-64.
[7]李丽君,金先龙.空间多点地震动模拟的并行计算方法及软件[J].系统仿真学报,2005,17(5):1 068-1 071.
[8]Kazuo Kashiyama,Katsuya Saitoh,Marek Behr,et al.Tezduyar.Parallel finite element methods for large-scale computation of storm surgesand tidal flows[J].International Journal for Numerical Methods in Fluids,1997,24(12):1 371-1 389.
[9]李友云,崔俊芝,尤勇,等.配电网络系统潮流计算的一种并行算法[J].数值计算与计算机应用,2004,3:233-240.
[10]杜鸿,赵越超,薛弘烨.实时集群计算技术在相控阵雷达中的研究[J].现代雷达,2004,26(6):22-25.
[11]Message Passing Interface Forum[EB/OL],http://www.mpi-forum.org/index.html
[12]MPICH2 Home Page[EB/OL],http://www-unix.mcs.anl.gov/mpi/mpich2/
[13]江春波,安晓谧,张庆海.二维浅水流动的有限元并行数值模拟[J].水利学报,2002(5):65-68.
[14]JIANG Chunbo,ZHANG Qinghai,An Xiaomi.River Flow Simulations Using Parallel Computing Techniques[A].Proceedings IAHR Interna-tional Conference[C].Beijing:2001.37-43.
[15]余欣,杨明,王敏,等.基于MPI的黄河下游二维水沙数学模型并行计算研究[J].人民黄河,2005,27(3):49-51.
[16]Download MPICH2[EB/OL],http://www-unix.mcs.anl.gov/mpi/mpich2/downloads/mpich2-1.0.2-1-win32-ia32.msi
[17]王光谦,刘家宏,李铁键.黄河数字流域模型原理[J].应用基础与工程科学学报,2005,13(1):1-8.
[18]刘家宏,王光谦,王开.大流域数字高程模型数据管理系统[J].清华大学学报(自然科学版),2004,44(12):1 646-1 649.
[19]刘家宏.黄河数字流域模型[D].北京:清华大学,2005.1-205.
[20]李铁键,王光谦,刘家宏.数字流域模型的河网编码方法[J].水科学进展,2006,17(5):658-664.
[2]Top500 List for June 2005[EB/OL],http://www.top500.org/lists/2005/06/
[3]田华,卞建春,颜宏.浅水波模式半拉格朗日方法的并行研究[J].应用气象学报,2004,15(4):417-426.
[4]袁金南,王在志,薛纪善.广州区域数值预报模式并行化计算[J].应用气象学报,2004,15(5):556-563.
[5]赵改善,包红林.集群计算技术及其在石油工业中的应用[J].石油物探,2001,40(3):118-126.
[6]赵改善,孔祥宁,王于静,等.64位集群计算平台波动方程叠前深度偏移的性能优化[J].勘探地球物理进展,2005,28(1):57-64.
[7]李丽君,金先龙.空间多点地震动模拟的并行计算方法及软件[J].系统仿真学报,2005,17(5):1 068-1 071.
[8]Kazuo Kashiyama,Katsuya Saitoh,Marek Behr,et al.Tezduyar.Parallel finite element methods for large-scale computation of storm surgesand tidal flows[J].International Journal for Numerical Methods in Fluids,1997,24(12):1 371-1 389.
[9]李友云,崔俊芝,尤勇,等.配电网络系统潮流计算的一种并行算法[J].数值计算与计算机应用,2004,3:233-240.
[10]杜鸿,赵越超,薛弘烨.实时集群计算技术在相控阵雷达中的研究[J].现代雷达,2004,26(6):22-25.
[11]Message Passing Interface Forum[EB/OL],http://www.mpi-forum.org/index.html
[12]MPICH2 Home Page[EB/OL],http://www-unix.mcs.anl.gov/mpi/mpich2/
[13]江春波,安晓谧,张庆海.二维浅水流动的有限元并行数值模拟[J].水利学报,2002(5):65-68.
[14]JIANG Chunbo,ZHANG Qinghai,An Xiaomi.River Flow Simulations Using Parallel Computing Techniques[A].Proceedings IAHR Interna-tional Conference[C].Beijing:2001.37-43.
[15]余欣,杨明,王敏,等.基于MPI的黄河下游二维水沙数学模型并行计算研究[J].人民黄河,2005,27(3):49-51.
[16]Download MPICH2[EB/OL],http://www-unix.mcs.anl.gov/mpi/mpich2/downloads/mpich2-1.0.2-1-win32-ia32.msi
[17]王光谦,刘家宏,李铁键.黄河数字流域模型原理[J].应用基础与工程科学学报,2005,13(1):1-8.
[18]刘家宏,王光谦,王开.大流域数字高程模型数据管理系统[J].清华大学学报(自然科学版),2004,44(12):1 646-1 649.
[19]刘家宏.黄河数字流域模型[D].北京:清华大学,2005.1-205.
[20]李铁键,王光谦,刘家宏.数字流域模型的河网编码方法[J].水科学进展,2006,17(5):658-664.
版权所有：© 2023 中国地质图书馆 中国地质调查局地学文献中心