Special Features:
GRIDSYSTEMS AND E-SCIENCE
by
Marco Laucelli
A little more than half a decade ago the appearance of the Web started a
reorganising process in our society. The fruits of that revolution are what is
usually called information society. In that moment only one technology, Web
technology, produced a bunch of new concepts related with several human
activities as communication, commerce and entertainment. The result, some
years later, is a set of concepts with a very similar technological subtract,
corresponding to the different opportunities, mainly commercial, that became
possible. This is the case of B2B (business to business), B2C (business to
consumer)...We could add, probably not without discussion, the famous Peer to
Peer in the long list that I don't want to detail. This proliferation in
business opportunities was possible because the technological transfer from
the scientific community to the society was clearly done. Once the technology
was understood, the opportunities easily appeared.
Nowadays, when the initial euphoria of Web technology is passed, a new
technology that promises to revolutionize our idea of communication networks
and computational resources is at hand. It is the Grid technology, and in a
few years it will allow us access in a distributed way not only to information
but also to Data Bases, computational resources, simulation, visualization and
mobile devices. I have sometimes heard comparisons of the web revolution and
the changes that Grid technology is supposed to produce. However, the process
looks like it is slower as supposed and the wanted revolution is being
transformed into an evolution.
From my point of view, the reasons why the Grid phenomenon is taking some
time
are mainly two: one technological and the other sociological.
In a first place, the Grid term is nowadays ambiguous. It is really a word
that is used to talk about several different concepts: MetaComputing, High
Throughput Computing, High Performance Computing, Peer To Peer (more
polemics), Distributed Data Bases, Beowulf, Parallelism... All these terms are
familiar in the Grid literature even if their relationship is, at least, not
very clear. The sociologic factor that has slowed down the Grid revolution is
ambition. The Grid projects that we are used to seeing are incredibly huge,
and are related with scientific global-level collaboration, where complex
computational processes are combined with huge amounts of data. This is what
is called e-science and, in the words of John Taylor, General Director of the
Britannic Research Councils and one of the instigators of the e-science in
that country, could be defined as: "In the future, e-Science will refer to the
large scale science that will increasingly be carried out through distributed
global collaborations enabled by the Internet. Typically, a feature of such
collaborative scientific enterprises is that they will require access to very
large data collections, very large scale computing resources and high
performance visualisation back to the individual user scientists."
So, it seems that what we are dealing with is the science of future, that
will
still have to wait for the complex management systems currently being
developed. In some years they will be of key importance to important
challenges as the new CERN accelerator, high scale meteorological predictions,
or the interactive processes for medical simulation. But, if we are talking
about the future, it means that we have to wait.
In the current situation we have several technologies with a confusing
common
name and a set of projects that apparently look not much related with the
current needs. At this stage it is not weird that Grid technology impact in
the corporate world is much less than expected. Even worse, in my opinion, is
the situation among the scientific community. Everyone has heard about Grid,
but not many use it in their day-to-day activity.
For a professional focused on Grid technology the previous thoughts should
lead to the following question: is it possible that Grid technology can be
used today and we don't need to wait for it so long? And, focusing on the
scientific world, is it possible to talk about e-science, even if it has to be
redefined, for the current and not future challenges?
Finding answers to these questions, and convincing several persons about
this,
has been my task the last months at GridSystems. To make Grid technology
accessible today we have to solve the difficulties that can cause its delay.
In the first place it has to be clearly defined what Grid is nowadays and what
it is suitable for, and second, try to solve the scientific computational
problems today (that exist). Let's continue our research with the future
problems, but let's pay attention to the tools available today and to the
problems that can already be solved.
At GridSystems we believe that with the current technology Grid has to be
understood as a distributed computational resources management system. Said in
a different way, the Grid technology available today allows getting the
maximum profit from the resources using the idle CPU cycles, memory...This
conviction has taken us to develop a system to manage the computational
resources and tasks in a distributed environment that we call InnerGrid. This
platform allows different computers in a network to be integrated, having any
operating system, to perform tasks that can be divided in independent
sub-tasks.
Which of the scientific computational processes can take advantage of a
platform like InnerGrid? Listing all of them would be impossible, but let's
think about some examples, such as materials analysis, Montecarlo simulation,
genetic sequences alignment... I'll come later to a more concrete example. We
are convinced that with InnerGrid it is possible to start a backbone of
computational infrastructure that allows establishing today useful
cooperation's among the scientific community and that will be the base for the
future e-science. For that, two additional requisites have to be met: easy to
use and implementation of the tools. InnerGrid has been designed and developed
with special attention to these features.
Universia, the Spanish & latin American Universities website portal
sponsored
by the SCH Bank, has bet for Grid technology as a present and future tool for
Spanish science, and has launched a project called Grid-Universia to give
access to this technology to the research groups in the Spanish universities.
The chosen tool for the project is InnerGrid, from GridSystems, which is also
taking care of the education and support of the users. During this project,
that will last till 2004, we have checked how Grid can help a lot to solve
current scientific problems. The projects that we have done in this frame,
about thirty, have allowed linking the computational resources of Faculties
and departments to establish cooperation frames for the researchers.
In some cases InnerGrid platforms have been installed with the goal to take
advantage of underused common resources. Some examples are the
Chemistry-Physics department at the University of Valencia that has decided to
use the PCs in classrooms PCs (about 200) to distribute Gaussian simulations;
the Faculty of Sciences oat the University of Oviedo has installed a Grid in
the students classrooms (about 50) to perform simulations for several research
teams (material physics, stochastic simulations, nuclear physics); the
University of Seville has a student's room as a common computational resource
for researchers (already been used for material simulations).
In other cases we have linked desktop computers to dedicated computers: the
Genetics department at the University of Valencia has integrated their
members'
computers and a PC farm in a Grid Platform to perform filogenetics analysis;
the Condensed Material Physics group of the University of Oviedo has done the
same for molecular dynamic simulations; the Signal Theory department of the
University Carlos III University in Madrid has integrated computers in their
laboratories with computational servers to distribute tasks with MatLab.
I could continue listing different cases of current uses of the Grid
technology by Spanish scientists in the frame of the Grid-Universia project,
but the ones already given are enough to demonstrate that e-science is
possible today. It is not the global e-science described by John Taylor, it is
not a big scale science, but it is a science that already uses information
technologies to new research possibilities. The local partnerships done thanks
to InnerGrid will continue adding new ones with the new needs evolving towards
an integrated fabric that will allow these every time more demanding
challenges. The current application of our Grid technology in the universitary
environment is not revolutionary, but it does allow us to recollect the first
fruits of this evolution. A similar strategy allows us to solve different
challenges in several corporate environments.
Marco Laucelli, R&D responsible of GridSystems
|
