A Web Interface for Meta-Heuristics Based Grid Schedulers
Fatos Xhafa
Department of Languages
and Informatics Systems
Technical University of Catalonia, Spain
Email: fatos@lsi.upc.edu
Joanna Kolodziej
Department of Mathematics
and Computer Science
University of Bielsko-Biala, Poland
Email: jkolodziej@ath.bielsko.pl
Marcin Bogda´ nski
Faculty of Mechanical Engineering
and Computer Science
University of Bielsko-Biala, Poland
Email: marcin.bogdanski@gmail.com
Abstract—The use of meta-heuristics for designing efficient
Grid schedulers is currently a common approach. One issue
related to Grid based schedulers is their evaluation under
different Grid configurations, such as dynamics of tasks and
machines, task arrival, scheduling policies, etc. In this paper
we present a web application that interfaces the final user with
several meta-heuristics based Grid schedulers. The application
interface facilities for each user the remote evaluation of
the different heuristics, the configuration of the schedulers
as well as the configuration of the Grid simulator under
which the schedulers are run. The simulation results and
traces are graphically represented and stored at the server
and can retrieved in different formats such as spreadsheet
form or pdf files. Historical executions are as well kept
enabling a full study of use cases for different types of Grid
schedulers. Thus, through this application the user can extract
useful knowledge about the behavior of different schedulers by
simulating realistic conditions of Grid system without needing
to install and configure any specific software.
Keywords-Web interface, Grid scheduling, Simulation, Meta-
heuristic.
I. I NTRODUCTION
Grid computing has emerged as a wide area distributed
platform for solving the large-scale problems in science,
engineering, etc. Computational Grid involves the combi-
nation of many computing resources into a network for
the execution of computational tasks. The resources are
distributed across multiple organizations, administrative do-
mains having their own access, usage policies and local
schedulers. The tasks scheduling and the effective manage-
ment of the resources in such systems are complex and
therefore, demands sophisticated tools for analyzing the
algorithms performances before applying them to the real
systems.
Simulation seems to be the most suitable way to analyze
scheduling algorithms in large-scale distributed dynamic
systems like Grid environment. It simplify the study of
schedulers performances and avoid the overhead of co-
ordination of the resources, which usually happens in the
real-life Grid scenarios. Simulation is also effective in work-
ing with very large problems that require involvement of a
large number of active users and resources, which is usually
very hard to coordinate and build for real-life approaches.
Moreover, in order to make useful the analysis of the
performance and parameters tuning of different scheduling
algorithms, a considerable number of independent runs is
needed to ensure significant statistical results, which can be
easily realized with Grid simulator.
Despite the work on the Grid simulation packages, there
are still few research works on development of full-featured
interfaces for Grid simulation packages. As a matter of fact,
current Grid simulation packages are run through command
line. In this work we propose a user-friendly Web interface
to remote users and benefit from easy use features of
the interface for running the HyperSim-G simulator. The
proposed Web application, namely WEBGRID
1
, facilitates
the evaluation of the different scheduling heuristics, the
configuration of the schedulers as well as the configuration
of the Grid simulator, under which the schedulers are run.
The simulation results and traces are graphically represented
and stored at the server and can retrieved in different formats
such as spreadsheet form or pdf files. Historical executions
are stored in the system, which enables a full study of use
cases for different types of Grid schedulers.
The rest of the paper is organized as follows. In Sec-
tion II we highlight the Grid scheduling problems types and
selected resolution methods. The HyperSim-G simulator as
well as its integration with various schedulers are briefly
described in Section III. In Section IV we present the main
aspects of the Web application architecture and show its
application in solving a specified instance of Grid scheduling
problem. Finally, we conclude this work in Section V.
II. SCHEDULING IN GRID SYSTEMS
A. Scheduling Problems
The purpose of the schedulers is to efficiently and opti-
mally allocate tasks originated by applications to a set of
available resources. In general, both tasks and resources
could be dynamically added/dropped to/from the system.
Scheduling in Grids remains a challenging NP-complete
global optimization problem because of the large-scale het-
erogeneous structure of the system, and the existence of
1
The application is available at the following Web page:
http://weboptserv.lsi.upc.edu/WEBGRID/
2010 International Conference on P2P, Parallel, Grid, Cloud and Internet Computing
978-0-7695-4237-9/10 $26.00 © 2010 IEEE
DOI 10.1109/3PGCIC.2010.67
405