The NeuroGrid Project
![](neurogrid-logo.gif)
Economic and On Demand
"Brain Activity Analysis" on the World Wide Grid Using Nimrod-G and Gridbus
Technologies
Introduction
The lack of computational power within an organization for analyzing scientific
data, and the distribution of knowledge (by scientists) and technologies
(advanced scientific devices) are two major problems commonly observed in
scientific disciplines. One such scientific discipline is brain science.
The analysis of brain activity data gathered from the MEG (Magnetoencephalography)
instrument is an important research topic in brain science since it helps
doctors in identifying symptom of diseases. The data needs to be analyzed
exhaustively to efficiently diagnose and analyze brain functions and this
exhaustive analysis needs large-scale processing resources. The emerging
Grid technologies that enable the sharing, selection, and aggregation of
geographically distributed resources can help in solving these problems.
However, application development, resource management and scheduling in Grid
environments is a complex undertaking.
We have developed a MEG data analysis system by leveraging Grid technologies,
primarily Nimrod-G and Gridbus. The wavelet analysis program has been parameterised
using the Nimrod-G parameter specification language. This application is
enabled for distributed processing on the Grid with with minimal software
engineering cost and development time. An application specific meta-job processing
plug-in module has been developed to enable the composition of a set of MEG
analysis jobs as a single coarse-grain processing job.
Furthermore, the Nimrod-G grid broker supports users’ quality-of-service
(QoS) requirements driven brain activity analysis application scheduling
on the Grid. In this system, we attempted to reduce analysis time (deadline)
and cost (budget) and to seamlessly integrate resources (computational, data,
and MEG instrument). Our evaluation results show that the system is highly
efficient in reducing the analysis time and cost. The results demonstrate
that grid technology is effective and promising for real-life medical and
scientific problems.
SC 2002 Conference HPC
Chellenge Entry
In the SC 2002 HPC challenge, we access 0.9 GB of brain activity data collected
using the 64-sensors MEG instrument (located in Osaka, Japan) for one hour
duration and perform on-demand brain activity analysis on globally distributed
computers. This problem generates 7257600 analysis jobs and expected to take
102 days, on a commodity computer with a PentiumIII/500MHz processor and
a 256MB memory. When such analysis is performed using our Grid brokering
system, users/doctors will be able to steer analysis activity depending on
QoS requirements (deadline, budget, optimization preference) and priorities.
We demonstrate adaptive and dynamic selection of resources at runtime depending
on their availability, capability, cost, data location, and users QoS requirements
for brain activity analysis. In this HPC challenge, we will make use of the
World-Wide Grid testbed resources located in Australia, Japan, Singapore,
Europe (Germany, Italy, Czech Republic, UK), and USA.
Software
- Cross Corelation and Wavelet Analysis Software.
- Gridbus Toolkit
: Visual Parameter Modelling Tools, Data Grid Scheduler, MetaJob Processing,
Grid Market Directory
- Grid Resource Broker (
Nimrod-G
)
- Grid Middleware (Globus
)
Computational Resources
Collaboration
Presentation/Exhibition
-
Brain Science on the Grid
, The 2nd PRAGMA
(Pacific Rim Applications and Grid Middleware Assembly) Workshop (PRAGMA-2),
July 10-11, 2002, Seoul, Korea
- Neuroscience on the Grid, IEEE Supercomputing Conference
(SC 2002
), November 16-22, 2002, Baltimore, USA.
- Rajkumar Buyya, Susumu Date, Yuko Mizuno-Matsumoto, Srikumar Venugopal1, and David Abramson, Composition of Distributed Brain Activity Analysis and its On-Demand Deployment on Global Grids, Technical Report, Grid Computing and Distributed Systems (GRIDS) Lab, Dept. of Computer Science and Software Engineering, The University of Melbourne, Australia.
Related Early Work using MPI (Explicit Parallelisation of Application)