Running Bag-of-Tasks Applications
on Computational Grids: The MyGrid Approach
Walfredo Cirne Daniel Paranhos Lauro Costa
Elizeu Santos-Neto Francisco Brasileiro Jacques Sauvé
Universidade Federal de Campina Grande
{walfredo,danielps,lauro,elizeu,fubica,jacques}@dsc.ufcg.edu.br
Fabrício A. B. Silva
UniSantos
fabricio@unisantos.br
Carla O. Barros
LNCC
osthoff@lncc.br
Cirano Silveira
Hewlett Packard
cirano.silveira@hp.com
Abstract
We here discuss how to run Bag-of-Tasks applications
on computational grids. Bag-of-Tasks applications (those
parallel applications whose tasks are independent) are both
relevant and amendable for execution on grids. However,
few users currently execute their Bag-of-Tasks applica-
tions on grids. We investigate the reason for this state of
affairs and introduce MyGrid, a system designed to over-
come the identified difficulties. MyGrid provides a simple,
complete and secure way for a user to run Bag-of-Tasks
applications on all resources she has access to. Besides
putting together a complete solution useful for real users,
MyGrid embeds two important research contributions to
grid computing. First, we introduce some simple working
environment abstractions that hide machine configuration
heterogeneity from the user. Second, we introduce Work
Queue with Replication (WQR), a scheduling heuristics
that attains good performance without relying on informa-
tion about the grid or the application, although consuming
a few more cycles. Note that not depending on information
makes WQR much easier to deploy in practice.
1. Introduction
Bag-of-Tasks (BoT) applications are those parallel ap-
plications whose tasks are independent of each other. De-
spite their simplicity, BoT applications are used in a vari-
ety of scenarios, including data mining, massive searches
(such as key breaking), parameter sweeps [1], simulations,
fractal calculations, computational biology [25], and com-
puter imaging [23]. Moreover, due to the independence of
their tasks, BoT applications can be successfully executed
over widely distributed computational grids, as has been
demonstrated by SETI@home [3]. In fact, one can argue
that BoT applications are the applications most suited for
computational grids, where communication can easily be-
come a bottleneck for tightly-coupled parallel applications.
However, few users of BoT applications are currently
using computational grids, despite the dramatic increase in
resources grids can potentially bring to bear for problem
resolution. We believe that this state of affairs is due to (i)
the complexities involved in using grid technology, and
(ii) the slow deployment of existing grid infrastructure.
Today, one must commit considerable effort to make an
application run efficiently on a grid. The user, who is ulti-
mately interested in getting the application’s results, sel-
dom has the training or the inclination to deal with the
level of detail needed to use current grid infrastructure.
Furthermore, the existing grid infrastructure is not ubiqui-
tously installed yet. Users often have access to resources
that are not grid-ready.
In this paper we present MyGrid, a system designed to
change this state of affairs. MyGrid aims to easily enable
the execution of BoT application on whatever resources
the user has available. MyGrid chooses a different trade-
off compared to existing grid infrastructure. It forfeits
supporting arbitrary applications in favor of supporting
only BoT applications (which are relevant and amenable to
execution on grids). By focusing on BoT applications,
MyGrid can be kept simple to use; simple enough to be a
solution for real users, who want to run their applications
today and don’t really care for the underlying grid support
they might use.
This is not to say, however, that MyGrid is a replace-
ment for existing grid infrastructure. MyGrid uses grid in-
frastructure whenever available. It simply does not depend
on it. MyGrid can be seen as a representative of the user in
the grid. It provides simple abstractions through which the
user can easily deal with the grid, abstracting away the
non-essential details (as we shall see in Section 4). It
schedules the application over whatever resources the user
has access to, whether this access is through some grid in-
frastructure (such as Globus [19]) or via simple remote
login (such as

). MyGrid’s scheduling solution is a par-
ticularly interesting contribution because it uses task repli-
cation to achieve good performance relying on no informa-
Proceedings of the 2003 International Conference on Parallel Processing (ICPP’03)
0190-3918/03 $ 17.00 © 2003 IEEE