1
Labs of the World, Unite!!!


Walfredo Cirne Francisco Brasileiro Nazareno Andrade
Lauro B. Costa Alisson Andrade
Universidade Federal de Campina Grande
Departamento de Sistemas e Computa ª o
Laborat rio de Sistemas Distribu dos
{walfredo, fubica, nazareno, lauro, aandrade}@dsc.ufcg.edu.br

Reynaldo Novaes Miranda Mowbray
Hewlett Packard
{reynaldo.novaes,miranda.mowbray}@hp.com


Abstract: eScience is rapidly changing the way we do research. As a result,
many research labs now need non-trivial computational power. Grid and volun-
tary computing are well-established solutions for this need. However, not all
labs can effectively benefit from these technologies. In particular, small and
medium research labs (which are the majority of the labs in the world) have a
hard time using these technologies as they demand high visibility projects
and/or high-qualified computer personnel. This paper describes OurGrid, a sys-
tem designed to fill this gap. OurGrid is an open, free-to-join, cooperative grid in
which labs donate their idle computational resources in exchange for accessing
other labs idle resources when needed. It relies on an incentive mechanism
that makes it in the best interest of participants to collaborate with the system,
employs a novel application scheduling technique that demands very little in-
formation, and uses virtual machines to isolate applications and thus provide
security. The vision is that OurGrid enables labs to combine their resources in a
massive worldwide computing platform. OurGrid is in production since Decem-
ber 2004. Any lab can join it by downloading its software from www.ourgrid.org.

Keywords: free-to-join grids; grid computing; grid scheduling; incentive to col-
laborate; peer-to-peer grids; sandboxing.

2
1. Introduction
The recent advances in computing and networking are changing the way we do scien-
tific research, a trend that has been dubbed eScience. Thanks to the power of computer-based
communication, research is now a much more collaborative endeavor. Moreover, computers
play an ever-increasing role in the process of scientific discovery. Data analysis without com-
puters sounds antediluvian. Simulation has joined theory and experimentation as the third sci-
entific methodology. As a result, many research labs now demand non-trivial computing capa-
bilities. Buying more computers is a natural answer to this demand. But compute demand
seems to be insatiable. No matter how much computing resource is available, it is common-
place to hear we could do more/better research if we had access to more computing power .
Computer scientists have long recognized this fact, and began to address it by provid-
ing a way to harvest the computing power going idle in one s lab or university [43]. This repre-
sented an important step forward, but has limited scale. Latter, this idea evolved into harvest-
ing the computing power going idle in the Internet [2] [3], in what became known as voluntary
computing. At about the same time, grid computing [11] [34] appeared with the enticing vision
of plug into the grid and solve your computational problem .
Voluntary computing has been able to deliver unprecedented computing power to some
applications. For example, at the beginning of March 2005, SETI@home had mustered more
than 2.2 million years of CPU time, from over 5.3 million users, spread across 226 countries
[57]. However, in order to benefit from voluntary computing, it is necessary to have a high visi-
bility project, set up a large control center to manage the volunteers, and put a lot of effort into
publicity to convince people to install the worker module. Naturally, being in a prestigious
University and having a qualified team for software development and system administration
can help a great deal towards this. Alas, these conditions do not hold for most research labs in
the world.
On the other hand, grid computing is turning from promise into reality. There are now a
few large Globus-based grids in production. With dozens of sites and thousands of computers,
CERN s LCG [17] is probably the best current example of what grids can achieve. However,
current grids are somewhat limited in scale [33] [51], not going beyond dozens of sites. More-
over, installing, configuring, and customizing Globus is not a trivial task, and currently requires
a highly skilled support team. Therefore, current grid solutions make excellent sense for doz-
ens of large labs that work together on similar problems. Again, that is not the case for most
labs around the world.