D-Grid – Towards E-Science in Germany
When Germany started its first Grid Initiative UNICORE
in the word Grid – as a description for transparent
usage of distributed resources by virtual organisations
– had not been coined yet. However, UNICORE and its
successor project UNICORE+ certainly marked the beginning
of a Grid initiative in Germany. Although it came up with
a product, it was clear at the end of the project that the
reach of UNICORE was too short. It only tackled the question
of accessing supercomputing resources, but there was no
whatsoever support for a full fledged e-science infrastructure.
This situation was addressed in the final UNICORE+ Review
Meeting in December 2002 and planning for a German e-science
initiative started immediately picking up on the many ideas
and other projects that had grown among a vibrant Grid community
in Germany over time. The code name for the initiative was
chosen to be D-Grid. The founders of the initiative are
faced with the following situation.
Situation in Germany
Grid computing is well established in Germany in a variety
of projects. German research groups are among the most actives
in European Grid Initiatives. Already in 1998 in the European
framework program 5 (FP5) a French-German pilot project
(METODIS) for the industrial usage of Grid tools was initiated.
A number of projects followed with German researchers partially
taking the lead. In FP6 German researchers were strongly
involved from the start both in projects building a European
research infrastructure and in using the Grid for complex
problem solving.
At the same time Grid research was conducted in a number
of communities. The most notable one currently is certainly
high energy physics (HEP). As part of the large hadron collider
(LHC) experiments of CERN the community has set up a concept
for the processing of the huge pile of data generated during
the experiments. At the Research Centre Karlsruhe (FZK)
a Grid Centre (GridKa) was established that acts as a tier
1 centre for the LHC. Other communities like climate and
weather research have established similar activities in
the field.
German researchers were also rather active in international
activities for a standardization of the Grid. Already the
foundation of a European Grid initiative was strongly driven
by German participants – although acting on a European
level with colleagues from many other countries. These activities
were intensified when the Global Grid Forum was established.
German researchers actively participated at all levels.
The importance of German Grid activities inside GGF
is certainly emphasized by the fact that GGF had its March
2004 meeting in Berlin.
A Vision for e-science
Modern information society is faced with a multiplicity
of possibilities for efficient communication and easy access
to large sets of data and information as well as high performance
computing systems. The opportunities to achieve better results
– both quantitatively and qualitatively – in
science and industrial development based on completely new
methods have increased dramatically. At the same time complexity
and thus difficulties in handling distributed and dynamic
system components have increased dramatically.
Virtualisation is therefore the key word for future Grid
systems. Hiding away the complexities of data – information
– and compute services and thus virtualising them
has to be the goal. Future Grid infrastructures will comprise
thousands of geographically distributed resources that are
interconnected by wide area networks. These resources will
autonomously supervise and reorganize themselves whenever
necessary. While providing reliable, fault tolerant and
ubiquitous services of high quality they will be transparent
to the user.
The goal of D-Grid is a bundling of Grid initiatives in
Germany to harness the potential of synergetic effects for
global, distributed and innovative collaborations in science
and industry. This requires the structuring both at a horizontal
and a vertical level as shown in Figure 1.
Figure 1:
The D-Grid e-science Framework
At the horizontal level synergy is achieved by bundling
and a common usage of middleware, services and e-science
methods across scientific communities which were so far
isolated. This brings about homogenisation and standardization
which allows harnessing synergy for scientific work. At
the lowest level we find resources like networks, computers,
programs, instruments and data archives that have to be
integrated to form a basic infrastructure.
At the vertical level an increase in efficiency for each
scientific discipline can be seen in the creation, bundling
and exploitation of services, resources and infrastructures.
Figure 1 shows how this can be achieved at each layer of
the conceptual view of the Grid.
Package of Measures
To achieve such an integration and exploitation of synergetic
effects a set of measures will have to be taken:
• Installation of a network
of competence in e-science: A well coordinated network of
competence based on centres of excellence is the essential
ingredient for a national e-science program. These centres
are crystallisation points of competence in the field of
Grid computing. A common head office coordinates the key
responsibilities which are
• Provision of resources
and know-how
• Development of application
scenarios for science and industry
• Research and development
in Grid systems
• Permanent integration of
user requirements
• Development of Grid software
components: Generic and standardized software is the basis
for Grid computing in science and industry. It has to be
developed in close co-operation with the international research
community and with industry
• Creation of a lasting
infrastructure: The core of such an infrastructure is formed
by existing and improved networks, computer systems and
data bases. This has to be complemented by services and
a middleware layer. Although based on existing national
infrastructures like the German research network DFN, the
German federal HPC centre resources or the UNICORE group,
this has to be extended into the international arena to
be competitive
• Setting up of a number
of pilot projects in e-science: Pilot projects do not only
help to demonstrate the benefit of a German Grid initiative.
They can also help to gradually lure more and more scientific
communities into the Grid arena. Thus a true e-science space
is created
Organisational Aspects
The German Grid community is based on self organisation.
At plenary meetings the main issues are discussed openly
and necessary decisions are made. Currently D-Grid is a
group of about 45 organisations. These come from universities,
research organisations and industry. Since D-Grid is an
open community any research group in Germany can join by
signing a memorandum of understanding and participating
in the coordination process. More information on the organization
and how to participate can be found at
http://d-grid.de.
Most of the information provided here was taken from this
webpage.
The community has established a steering committee in
May 2003 for the preparation of a research program and to
prepare organisational concepts. Its main duties are:
• To work out a scientific
vision and goals for a German Grid research program together
with leading scientists in the field
•
To work out a concrete R&D program focussing on infrastructures,
Grid software and applications
•
To act as a liaison for the federal Ministry of Science
for the topics mentioned above
•
To integrate various scientific fields and communities
• To work out a memorandum
of understanding and lay the foundation for a work program
for a German Grid and e-science initiative.
The steering committee has proposed a matrix structure
for the further work on a research program. This matrix
structure can also be seen as the basis for the further
organisation of D-Grid. At the horizontal level a set of
working groups (WG) was established:
• Operations
• Networks
• Middleware
• Autonomic Computing/Management
• Data/Information
At the vertical level individual communities were asked
to collect and formulate their requirements as well as visions
for the future. Communities can centre on an application
field – as is the case for the High Energy Physics
(HEP) community or the Climate and Weather Research community.
They can also be communities of providers like supercomputer
centres or network providers.
These working groups and communities have created a work
program that is finalised in June 2004. Based on this work
D-Grid expects to be able to launch operation in early 2005.
 top
|
index
prev