Information Technology Reference
In-Depth Information
of different combination of materials for a new funnel in sandwich technology
for fire safety on ships. The test of the solution showed that the computing time
for a required simulation of 60 minutes real time of fire that needs approxi-
mately one month of computing time on a single PC can be reduced to one day
on a 32-node cluster. The experiment also showed the technical feasibility of
sourcing external computing resources.
The two experiments had several common characteristics and problem areas:
• Tasks that require high computing power and are crucial part of the core activi-
ties of the organizations involved - treatment calculation in hospitals and collab-
orative product engineering at the shipyard.
• Lack of own infrastructure and knowledge that can support tasks requiring high
performance computing.
• Support of tasks with demand for high performance computing with applica-
tions that are not designed for use in a Grid environment.
Similar characteristics have been observed also with other BEinGRID Business
Experiments for example in the BE addressing computational fluid dynamics simu-
lation in the automotive industry (BE01 2009) or the BE considering engineering
and business processes in metal forming in manufacturing (BE08 2009) (for more
examples see BEinGRID Booklet 2009). All experiments were able to provide a
proof that computing intensive tasks can be accelerated considerably by using Grid
computing either by sourcing external computing power through utility computing
or by creating internal clusters.
However, the Business Experiments revealed also that application of utility
computing or external high computing resources requires considerable changes in
the way how information and communication technology is handled in organisa-
tions.
First of all, the experiments showed that in general usage of HPC, or IaaS in
case of Clouds, requires Grid-enabled applications. This means that the application
deployed on a Grid needs to be able to parallelize and distribute jobs among avail-
able resources - independent of the fact if the resources are internal or external.
Availability of utility computing on the market per se is somehow irrelevant,
if there are no applications that can take advantage of it. Through the Business
Experiments it became evident that in order to be able to use Grid computing, an
initial investment is necessary for the adjustment, i.e. Grid-enablement, of existing
software applications. This initial investment needs to be considered by the external
Independent Software Provider (ISP) or by the company itself in case of own appli-
cations. The Grid-enabled application can then be deployed either in a SaaS manner
or on company internal Grids (see the BEinEIMRT Business Experiment). Both
approaches require substantial organizational and IT governance changes.
The usage based on SaaS implies that internal resources and licenses for the
application are combined with a SaaS pay-per-use licenses and usage. The access of
the application in a SaaS manner requires new contracts and license agreements with
the software vendor providing it (see for example Stanoevska-Slabeva et al. 2008).
