CROWN: A Service Grid Middleware for e-Science - Grid Computing: Infrastructure, Service, and Applications - page 26

the performance of our implementation through comprehensive experi-
mental studies. CROWN Security aims to satisfy some security require-
ments of dynamic distributed resources sharing and integration, but much
work remains to be done.

1.5
CROWN is now becoming one of the important e-science infrastructures
in China. We have developed and deployed a series of applications from
different disciplines, which include Advanced Regional Eta-Coordinate
Numerical Prediction Model (AREM), Massive Multimedia Data
Processing Platform (MDP), gViz for visualizing the temperature i eld of
blood l ow, Scientii c Data Grid (SDG), and Digital Sky Survey Retrieval
(DSSR) for virtual observation. These applications have been used as test
cases to verify the technologies in CROWN.
AREM uses a grid as a tool to study and rei ne the numerical prediction
models of weather and climate. Several numerical models are worked out
by meteorologists during their research and prediction work. Typically
these models use the raw weather data from a national meteorology
authority as inputs and simulate the weather transformation according to
the laws of atmospheric physics and l uid dynamics. The output can be
used as a prediction result of future weather. The simulations are all based
on complex numerical calculations and need large quantities of computing
power and storage capacities. By using the resource organization and job
scheduling technologies provided by CROWN, we successfully developed
the AREM research system. We encapsulated the Fortran complier, visuali-
zation tools (GrADS), and the simulation framework of AREM as services,
and a unii ed raw weather data center is also deployed. Meteorologists can
submit simulation jobs to the system and rei ne their numerical models
according to the results. Since the jobs are executed using the resources pro-
vided by the CROWN testbed, the execution procedure can be parallel, the
execution time can be greatly reduced, and the efi ciency of weather system
research and prediction model rei nement can be improved.
Large amounts of storage capability and computing power are needed
when performing multimedia data processing, such as content recognition
of voice or video. Traditionally a central processing model is applied and
pieces of data are collected and processed in a single point. When the input
data are increased, this method provides little scalability, especially for real-
time applications. We combined the service grid technologies with massive
data processing and implemented the MDP platform for multimedia data
processing. MDP has been deployed into CROWN and has provided service
since 2005. We encapsulated the related algorithms into services and deployed

Testbed and Applications of CROWN