The Realization of GRAPES Model on the
China Meteorological Application Grid Environment
Yang Xuesheng1
Zhang Weimin2
Liu Yongzhu1 Chen Dehui1
Liu Cancan2
Sun Jian1
1
(State Key Laboratory of Severe Weather, Chinese Academy for Meteorological Sciences, Beijing
100081, P,RChina)
yangxs@cma.gov.cn
2
(Parallel and Distributed Processing for National Laboratory, Changsha)
wmzang104@tom.com
Based on the GRAPES (Global/Regional Assimilation and PrEdiction System)
and grid technique, the China Meteorological Application Grid (CMAG) has been
established. It mainly consists of CMAG portal, GridWeather user interface system,
source code management system (CVSexplorer), GRAPES assimilation and
prediction system, and visualization system, etc. GRAPES was a unified model
system which developed by CAMS both for global and regional weather prediction
applications.
And CMAG portal includes Resource Access Portal (RAP), NWP Products
Browse Portal (PAP) and Remote Collaborative Portal (RCP). RAP provides an
approach to access geographically distributed computing resources. PAP disseminates
NWP services to the public and professional clients. While RCP offers a universal
platform for scientists to develop NWP modules, to share source codes, to register
modules within the grid and to perform NWP experiments.
Authenticated users can modify the GRAPES workflow, the forecast domain,
time step, dynamical scheme, physical parameterization schemes according to his
requirements via GridWeather interface.
A grid platform has been established, computing resources such as SunWay
cluster 48I and Oscar cluster 32 at CAMS, SunWay 32I and 32P at NMIC, Dawning
cluster at Guangzhou Regional Meteorological Center and Guangxi Meteorological
Bureau, Ocean cluster at Shanghai Typhoon Institute and Dawning 4000A at Shanghai
Super-computing Center have been aggregated. GRAPES system is running everyday
within this platform.
With the implementation of CMAG into the existed operational NWP system and
research environment, substantial benefits can be achieved, for example, it can
aggregate the geographically distributed computing resources and give end users
remote access to the computing, data and storage resources they need, can establish a
virtual research center and a uniform platform to make good use of expertise to
develop the new NWP system, can accelerate the procedure from research to
operation and can help the under-developing regions to establish the NWP system.
The realization of CAMG provides a remote collaborative research environment
for the geographical distributed scientists to make contributions to the development of
the GRAPES.
Researchers
CMA:China Meteorological Administration
Job Submission
Visualization
SSC: Shanghai Super-Computing Center
STI: Shanghai Typhoon Institute
CMAGPortal
Grid Resoures And Platform
Grid Resources in CMA
DATA
SUNWAY32P
SUNWAY32I
SUNWAY48I
OSCAR
GRMC DawninCluster
FTP,GridFTP
SSC Dawning 4000A
GuangXi Dawning
Data Transfer
STI Ocean Cluster
NWP Products
NUDT Yinhe Cluster
Wuhan IBM RS6000
Vice Portal
Vice Portal
visualization
Researchers
Visualization
Fig. 1 Flowchart of Remote Collaborative Portal
Grid based CMAG job workflow
Job log
CMAGPortal
forecaster
Data
Resource
by
selection
Script edition
job submission by
transfer
Gram Monitoring
GridFTP
Result transfer
to local vice portal
Forecast analysis
Fig. 2 Flowchart of operational GRAPES model system running on the CMAG environment
IE browser
Tomcat5.28 Application Server
CVSManager
CodeExplorer
GridUser Portlets Forecast Portlets
ProjectManager
PrecipitionObserved
ElementForecast
CityForecast
RainForecast
GridUserMap
CredentialDelete
CredentialSign
Base Portlets GridWeather portlets
CredentialRequest
ResourceBrowser
FileTransfer
FileManager
JobPosttreatment
JobMonitoring
JobSubmission
CredentialRecieve
PortletApplication
ProfileManager
LoginPortlet
LayoutManager
CMAGPortal Page
CVSExplorer
GridSphere Container
Mysql
Ganglia
Database Server
GridServices
Myproxy
GRAPES service
Grid Software(GT2,GT3,GT4,GOS)
Job Scheduler (OpenPBS,LSF,Conder)
Grid Resources(Cluster, expensive instrument)
Fig 3 Framework of CMAG
CVS Server
Computing
Location
resources
Oscar
CAMS
Grid
NWP
Job
Software
software
scheduling
installed
installed
GT4
Security
CMA
CA GRAPES
OpenPBS
CA GRAPES
OpenPBS
Center
Sunway48I
CAMS
GT4
CMA
distributed
package
Sunway32I
NMIC
GT4
CMA
CA GRAPES
OpenPBS
distributed
package
Sunway32P
NMIC
GT4
CMA
CA GRAPES
OpenPBS
distributed
package
Ocean
STI
GT4
CMA
CA GRAPES
OpenPBS
distributed
package
Dawning
GRMC
GT4
CMA
CA GRAPES
OpenPBS
distributed
package
Dawning
GXMB
GT4
CMA
CA GRAPES
OpenPBS
distributed
package
PC
WHSI
GT4
CMA
CA GRAPES
No
distributed
package
Dawning4000A SSC
GT3
SSC
CA GRAPES
distributed
package
Table 1 Computing platform on CMAG
LSF