Grid Computing at IHS

Bunti Ogor I., Lippold F., Ruprecht A.

Introduction

Grid computing is a growing computing technology which provides the ability to perform
computational tasks using numerous networked computers, gathering them in one virtual
computer architecture, i.e. grid, that is able to distribute the computational process across
this parallel infrastructure. Thus, grids provide the ability to perform computations on very
large data sets according to the size and availability of the computational resources in a
certain grid. It enables the better exploitation of the available computing power connected into
a network.

The basic idea of grid computing is to offer a model for solving excessively large
computational problems by using the idle resources of large number of computers treated as a
virtual cluster embedded in a distributed telecommunications infrastructure [1]. Grid is often
understood as a synonym for the term of cluster computing. Generally the main difference
between these two models is that cluster is a single set of uniform connected computers
placed at one location, while grid is composed of many worldwide placed clusters or other
kind of computers together with other kind of network or storage resources. Grid computing
focuses on the ability to support computation across administrative domains apart from
traditional computer clusters or classical distributed computing. It involves sharing
heterogeneous computational resources, based on different platforms and hardware/software
systems, which belong to different administrative domains.

One of the first and the most famous grid computing projects is SETI@home (Search for
Extraterrestrial Intelligence) [2], launched at UC Berkeley. It started in 1999 with the aim to
process massive radio telescope data. Specific for SETI project is that the computational
program runs as a screensaver on users personal computer (while the computer is idle) which
processes small pieces of overall data returning them after the process back to the main server
over Internet network.

Inspired by SETI project many other grid computing projects have been established with
application areas in protein, drugs, earthquake and astrophysical simulation as well as
climate/weather modelling and prediction. Engineering applications has lately started to
consider grid computing as a promising computing model enabling more detailed and quicker
engineering simulation. Although the grid concept has started as a scientific tool, nowadays
there are numerous computing vendors actively involved in commercialisation of the grid
conceptual framework.

One term very close connected to the grid computing is Web service. According to [3], a
Web service is a software system designed to support interoperable computer interactions
over a network using open standards and protocols. Thus, software applications written in
various programming languages and running on various platforms can use web services to
exchange data over networks in a similar way such as inter process communication on a
single computer. In this way, grid computing can be available and accessed as a grid service,
or kind of web service with special functionalities.
Flow simulation as application for Grid Computing

For thirty years Computational Fluid Dynamics (CFD) has been more and more spreading,
becoming a kind of standard tool for the analysis of complex flows. It has managed to reduce
the time and costs for design and optimisation of complex machines, e.g. water turbines or
pumps. CFD is especially used in the planning and design process of new hydro power plants.
It also enables easier improvement and increase of efficiency of already existing plants.
Nowadays the power plant design process requires the flow simulations, usually three
dimensional, of the complete plant system. These simulations are very complex and demand
high computational power. This computational power at low costs has been one of the
limiting factors for performing such simulations. Software licenses also contribute additional
costs.

Developing grid computing can be one promising solution to promote and utilise simulations
of complex systems. It is expected to have a major effect on large-scale and data intensive
computing applications in the field of technology and engineering in near future. Grid
computing enables the virtualization of distributed computing and data resources granting
users and applications access to numerous computing capabilities at runtime depending on
their availability, capability, performance, costs and quality of service.

D-Grid and InGrid projects and IHS contribution

Since September 2005 five community projects and the D-Grid integration project (DGI) have
started within the D-Grid consortium to build a sustainable grid infrastructure, and are thus
establishing the methods of e-Science in the German scientific community [4]. All
community projects will develop, together with the integration project, a general grid
infrastructure that will be available for all German scientists (see Fig. 1). The main idea of the
German e-Science program is to cover the domains of grid computing, e-Learning and
knowledge management. All projects participating in the D-Grid initiative are funded by the
German Federal Ministry of Education and Research.

Figure 1. Structure of D-Grid integration project [4]

The community project InGrid (Innovative Grid Developments for Engineering Applications)
is a part of this D-Grid initiative. InGrid is based on five prototype applications in the fields
of coupled multiscale problems, coupled multidisciplinary simulations and distributed
simulation based optimization [5]. Its main goal is the development of adaptive, scalable
process models and runtime environments to enable modelling, optimization, and simulation
of engineering applications from areas such as foundry technology, forming, groundwater
flows, turbine simulation or fluid-structure interaction. The aim is to give the user simulation-
based tools on different levels of complexity.
Eight German research and industrial institutions are participating in InGrid project. This
project is seen as a close collaboration between basic research as well as application oriented
research to establish a grid-based, networked hardware and software infrastructure for the
support of scientific engineering work. The development of a grid-based "computational
engineering community" demonstrates technologically advanced application of engineering
work in research as well as in operational innovation management.

It is planned that project results could be used on three levels, in order to interlink research,
development and production strategies [5]:
by direct exertion of influence on engineering applications at the involved universities,
by transfer of the developed methods and tools into the German economy and
by implementation of the developed software components into German and European
e-Science infrastructures.

The Institute of Fluid Mechanics and Hydraulic Machinery (IHS) is one of the involved
partners in InGrid project. Its task inside InGrid is at first the conceptual planning of one
simulation grid service based on the in-house CFD simulation code FENFLOSS (Finite
Element based Numerical FLOw Simulation System), as well as providing grid services for
geometry modelling of hydro power plants construction units and hydraulic machines in
general. The virtual prototype test bed, developed and integrated into the COVISE
visualisation software developed at High Performance Computing Center Stuttgart (HLRS)
(for more details refer to [6]), serves as a basis. Both software, FENFLOSS and COVISE,
have to be additionally extended for full grid computing support and functionality, especially
interaction with middleware used in grid computing (intended is the use of Globus
middleware [7]).

Figure 2. FENFLOSS in HLRS test grid environment (figure source [6])

Furthermore, the IHS provides various important applications, from flow simulations, flow-
structure interaction to optimisation. They can be classified as a kind of multi physics
problems and are usually of large scale. The optimal simulation of these problems often
requires high computational power or various computer architectures and resources. Hence,
combining different hardware architectures and software solutions, newest grid technology
provides a key to tackle practice relevant engineering problems.

The objective of this project is to offer this built-on and tested grid environment with all its
components as a kind of web service to those business establishments whose limited financial
capabilities do not allow them the usage of high performance computing.

As a conclusion it can be pointed out that the InGrid project develops new scientific approach
for the central question: how can grid technology support engineering in the future? It builds
on the idea that the abilities of man and machine have to be engaged in a most synergetic way
and that in particular adequately interlaced grid based engineering has a substantial
significance for future technical products and product development processes.

References

1. Joseph J., Fellenstein C., Introduction to Grid Computing, Prentice Hall, 2004
2. http://setiathome.berkeley.edu
3. W3C Web Services Activity home page (www.w3.org)
4. www.d-grid.de
5. www.ingrid-info.de
6. www.hlrs.de
7. www.globus.org