HydroMP:基于云计算的水动力学建模及计算服务平台

删除或更新信息，请邮件至freekaoyan#163.com(#换成@)

清华大学辅仁网/2017-07-07

刘荣华¹,魏加华¹(

),翁燕章¹,王光谦¹,唐爽²

2. 中国石油勘探开发研究院,北京 100083

HydroMP: A cloud computing based platform for hydraulic modeling and simulation service

Ronghua LIU¹,Jiahua WEI¹(

),Yanzhang WENG¹,Guangqian WANG¹,Shuang TANG²

1. State Key Laboratory of Hydroscience and Engineering, Tsinghua University, Beijing 100084, China
2. PetroChina Research Institute of Petroleum Exploration & Development, Beijing 100083, China

摘要:

HTML
输出: BibTeX | EndNote (RIS) 背景资料

文章导读

摘要为解决传统开发水利模拟服务模式中存在的计算精度与效率矛盾及模型接口不统一等技术瓶颈,构建了基于云计算的水动力学模拟服务平台,并提出云计算模式下一维水动力学模型的集成方法和调度模式,通过对模拟模型数据接口、模型调用、模型交互及处理、计算资源分配等方面进行封装,采用Web Service、 XML (Extensible Markup Language)及OpenMI (Open Model Interface)等技术实现了云计算平台下模型的调用、模型与平台间的数据交换和资源动态分配。以南水北调中线工程水力响应分析为例,应用开发的云计算水动力学模型进行了多方案、多层次的模拟计算,在计算效率、兼容性等方面具有良好的效果。研究表明:云计算模式下,可实现水动力学模型在广域互联网上的多客户端、多用户、多方案并发计算,采用云端计算和存储模式降低了高性能计算的门槛,提高了模型复用率和计算资源利用效率。

关键词 ：云计算,水动力学模型,模型集成,Web Service

Abstract：Computing accuracy in large CFD problems is restricted by the computation cost. A hydraulic modelling platform (HydroMP) was constructed using computing cloud to reduce costs and to provide a uniform interface between the dataset and the model for hydraulic models. This paper gives a generic integration method for hydraulic models and a scheduling mode for the cloud computing. The system includes the models and computing resources, including the data interface, model calls and interactions and computations of resource allocation using Web Services, XML(Extensible Markup Language) and OpenMI(Open Model Interface). This design assures that the cloud computing platform can conveniently call the models, exchange data between the servers and the clients, and dynamically allocate the computing resources including multiple cores and multiple nodes. The platform with the integrated flow model is used to analyze the hydraulic response to a change in the discharge of the South-to-North Water Diversion Project, which is a multi-client, multi-level and multi-scheme simulation. The results indicate that this model improves the computational efficiency and compatibility. This hydraulic computing model based on cloud computing theory allows multiple schemes be solved simultaneously using multiple clients in a WAN Internet and multiple users. The cloud computing model lowers the threshold for using HPC resources and improves the utilization efficiency of computation resources.

Key words：cloud computing hydraulic models model-integration Web Services

收稿日期: 2013-09-22 出版日期: 2015-09-03

ZTFLH:

基金资助:“十二五”国家科技支撑计划资助项目(2013BAB05B03,2013BAB05B05);水利部公益性行业科研专项(201201050);水沙科学与水利水电工程国家重点实验室自主课题(2012-KY-05)

引用本文:

刘荣华, 魏加华, 翁燕章, 王光谦, 唐爽. HydroMP:基于云计算的水动力学建模及计算服务平台[J]. 清华大学学报（自然科学版）, 2014, 54(5): 575-583.
Ronghua LIU, Jiahua WEI, Yanzhang WENG, Guangqian WANG, Shuang TANG. HydroMP: A cloud computing based platform for hydraulic modeling and simulation service. Journal of Tsinghua University(Science and Technology), 2014, 54(5): 575-583.

链接本文:

http://jst.tsinghuajournals.com/CN/或 http://jst.tsinghuajournals.com/CN/Y2014/V54/I5/575

图表:

HydroMP的总体框架

一维水动力学模型的数据结构

类的引用关系

类创建及引用流程

基于管道及MPI的并行集成模式

运行阶段	调用过程	参数
初始化	Initialize()	Simulation
初始化	GetComptime()	Simulation
创建管道	Validate()	Simulation
创建管道	Create_Pipe()	pipeID
数据交换	SimuConvert2ByteStr()	Simulation
数据传输	SimustrTransfer()	SimuByteString
数值计算	PerformTimestep()	TimeStepNum
	GetResult()	SID, Timespan
	SaveState()	State
	ClearState()	StateID
完成	Resultupload()	ResultArray
完成	Releasecache()	SID

可交互集成阶段及调用的过程

工况	渠首通过流量占比 /%			变化用时/min			方案计算范围
工况	初始时刻		终止时刻		渠首流量变化	闸门开度	分水口流量
1	70	50		10	10	10	全线
2	70	10		30	30	30	全线
3	70	80		10	10	10	全线
4	10	70		30	30	30	全线
5	70	10		30	30	30	黄河以南段和黄河以北段

总干渠输水流量变化工况组合表

方案提交界面

项目	性能
头节点及服务器	16核(2.4 GHz), 156 GB内存
计算节点	20个计算节点,单计算节点含24核(2.4 GHz)、 32 GB内存。

云计算服务器配置

序号	方案名	计算时长/h	所选模型	模型类型	计算时间/s	原计算时间/s	加速比
1	工况1	2 000	JPWSPC-SC	串行集成	415	400	—
2	工况2	2 000	JPWSPC-SC	串行集成	400	398	—
3	工况3	2 000	JPWSPC-MPI	并行集成	56	390	6.9
4	工况4	2 000	JPWSPC-MPI	并行集成	54	425	7.87
5	工况5	2 000	JPWSPC-MPI	并行集成	61	388	6.26

不同方案、不同模型的计算速度对比表

参考文献:

[1]	Launder B E, Spalding D B. The numerical computation of turbulent flows[J]. Computer Methods in Applied Mechanics and Engineering, 1974, 3(2): 269-289.
[2]	张刚,解建仓,罗军刚. 洪水预报模型组件化及应用[J], 水利学报, 2011, 42(12): 1479-1486. ZHANG Gang, XIE Jiancang, LUO Jungang. Componentized model of flood forecast and its application[J]. Journal of Hydraulic Engineering, 2011, 42(12): 1479-1486. (in Chinese)
[3]	Gregersen J B, Gijsbers P J A, Westen S J P. OpenMI: Open modelling interface[J]. Journal of Hydroinformatics, 2007, 9(3): 175-191.
[4]	Bulatewicz T, Allen A, Peterson J M, et al.The simple script wrapper for OpenMI: Enabling interdisciplinary modelling studies[J]. Environmental Modelling & Software, 2013, 39: 283-294.
[5]	Sun A. Enabling collaborative decision-making in watershed management using cloud-computing services[J]. Environmental Modelling & Software, 2013, 41: 93-97.
[6]	Burger C M, Kollet S, Schumacher J, et al.Introduction of a web service for cloud computing with the integrated hydrologic simulation platform ParFlow[J]. Computers & Geosciences, 2012, 48: 334-336.
[7]	Ari I, Muhtaroglu N. Design and implementation of a cloud computing service for finite element analysis [J]. Advances in Engineering Software, 2013, 60-61: 122-135.
[8]	Horsburgh J S, Tarboton D G, Maidment D R, et al.A relational model for environmental and water resources data[J]. Water Resources Research, 2008, 44(5): W05406.
[9]	Horsburgh J S, Tarboton D G, Schreuders K A T, et al. HydroServer: A platform for publishing space-time hydrologic datasets [Z/OL]. (2013-06-09). http://his.cuahsi.org/documents/JSH_SpringRunoff_HydroServer_2012.pdf.
[10]	Ames D P, Horsburgh J S, Cao Y, et al.HydroDesktop: Web services-based software for hydrologic data discovery, download, visualization, and analysis[J]. Environmental Modelling & Software, 2012, 37: 146-156.
[11]	Booch G, Maksimchuk R A, Engle M W, et al.Object-Oriented Analysis and Design with Applications[M]. Boston, USA: Addison Wesley, 2007.
[12]	周振红,任慧,杜丽平. Fortran DLL 组件集成到. NET平台(一)[J]. 武汉大学学报: 工学版, 2005, 38(4): 100-104. ZHOU Zhenhong, REN Hui, DU Liping. Fortran DLL component being integrated into Microsoft. NET framework[J]. Engineering Journal of Wuhan University, 2005, 38(4): 100-104. (in Chinese)
[13]	朱仕杰,南卓铜,陈昊,等. 基于Web Service的在线水文模型服务研究[J]. 遥感技术与应用, 2010, 25(6): 853-859. ZHU Shijie, NAN Zhuotong, CHEN Hao, et al.Online hydrological model service using web service[J]. Remote Sensing Technology and Application, 2010, 25(6): 853-859. (in Chinese)
[14]	ZHU Dejun, CHEN Yongcan, WANG Zhiyong, et al.Simple,robust, and efficient algorithm for gradually varied subcritical flow simulation in general channel networks[J]. Journal of Hydraulic Engineering—ASCE, 2011, 137(7): 766-774.

[1]	王志华, 庞海波, 李占波. 一种适用于Hadoop云平台的访问控制方案[J]. 清华大学学报（自然科学版）, 2014, 54(1): 53-59.