419 GANESH
GANESH: GRID APPLICATION
MANAGEMENT AND ENHANCED
SCHEDULING
H. S. Bhatt
R. M. Patel
H. J. Kotecha
V. H. Patel
A. Dasgupta
SAG/SITAA, SPACE APPLICATIONS CENTRE,
INDIAN SPACE RESEARCH ORGANISATION,
AHMEDABAD – 380 015, INDIA
(HARESH@SAC.ISRO.ORG)
Abstract
Grid computing is emerging as the next generation com-
puting environment. Various meta-computing and distributed
computing environments are becoming grid compliant
with the aim of achieving interoperability. We have devel-
oped a web-based development environment for distrib-
uted computing that supports heterogeneous resources
and provides ease of use. We present herewith GANESH
that achieves grid compliance using Globus and also
extends features for added value. This paper describes
GANESH architecture and its services, and explains its
grid compliance and extended features. We also explain
in this paper how and where it is interoperable with other
environments.
Key words: grid computing, high performance computing,
scheduler, workflow
1 Introduction
Diverse research advances in the areas of distributed com-
puting, parallel processing, cluster computing and meta-
computing have demanded a need for convergence. Peo-
ple across the world share the resources to achieve com-
mon objectives and are treated as the virtual organization
of a community by the grid technology (Foster, Kessel-
man and Tuecke 2001). Grid technology researchers aim
to provide seamless access to geographically distributed
resources and resource management environments (Cza-
jkowski et al. 1998). This can be achieved by defining
standards, protocols and services that provide interopera-
bility across heterogeneous resources (Foster et al. 2002;
Tuecke et al. n.d.). The Globus core team has significantly
contributed by providing toolkits (Foster and Kesselman
1997; von Laszewski et al. 2001).
The distributed and meta-computing environments
(Henderson and Tweten 1996; IBM 1996; Buyya, Abram-
son and Giddy 2000; Frey et al. 2001; Jackson n.d.;
OpenPBS
1
) are now becoming grid aware by using and
extending the grid protocols for security, resource man-
agement, job allocation, job monitoring, and so forth.
DEDIP is a distributed computing environment that
provides facilities for development, building and opera-
tionalization of image processing applications, aiming at
ease of use (Bhatt, Singh and Aggarwal 2002). It supports
VAX/VMS as the master and Sun Solaris or XENIX (Unix
flavor) as the slave. It was extended further as WebDEDIP
to make it web aware by providing a roaming profile to
the end user for accessing all its functionalities (Bhatt and
Aggarwal 2000). It works on JVM. It provides browser-
based GUI for (1) application configuration, (2) applica-
tion building, (3) application execution, and (4) applica-
tion monitoring. WebDEDIP redirects all the errors and
output of all processes to different files on the web server.
GUI enables the user to view them through a browser.
Special care is taken while providing the GUI to support
the operation center. The operation center provides services
to scientists to execute their programs on different datasets
and submit final results, relieving them from the burden of
performing repeated executions. It supports automated file
transfer among the nodes as required by the applications’
components. The user can manually provide the data trans-
fer information statically through the GUI. For dynamic
data transfer, it supports a library routine for generating this
information dynamically during the application execution.
The user has to simply call the routine with the required
information. The routine will generate the required file in a
predefined format. It supports multiple simultaneous appli-
cations. It will not only execute the tasks configured by each
application but also provides various status displays to the
operator to monitor and control his session.
Although WebDEDIP was developed for image process-
ing applications, it is quite generic in nature. It is equally
The International Journal of High Performance Computing Applications,
Volume 21, No. 4, Winter 2007, pp. 419–428
DOI: 10.1177/1094342007083777
© 2007 SAGE Publications Los Angeles, London, New Delhi and Singapore
Figure 5 appears in color online: http://hpc.sagepub.com