Porting HPC Applications to Grids and Clouds - Cloud, Grid and High Performance Computing: Emerging Applications - page 34

CONCLUSIONS: TEN RULES FOR
BUILDING SUSTAINABLE GRID
AND CLOUD E-INFRASTRUCTURES
FOR HPC APPLICATIONS

(Walker, 2008). The performance limitations in
clouds are exhibited mainly by parallel applica-
tions with tightly-coupled, data-intensive inter-
process communication, running on hundreds or
even thousands of processor cores.
The good news is, however, that many HPC
applications do not require high bandwidth and
low latency. Examples are parameter studies
(sweeps) often seen in science and engineering,
with one and the same application executed for a
spectrum of parameters, resulting in many inde-
pendent jobs, such as analyzing the data from a
particle physics collider, identifying the solution
parameter in numerical optimization, ensemble
runs to quantify climate model uncertainties,
identifying potential drug targets via screening a
database of ligand structures, studying economic
model sensitivity to parameters, simulating flow
around an airplane wing with different angels of
attach, and analyzing different materials and their
resistance in crash tests, to name just a few.
HPC needs Grids and clouds . According to
the DEISA Extreme Computing Initiative (DECI,
2010), there are plenty of complex grand chal-
lenge science and engineering applications that
can only run effectively on the largest and most
expensive supercomputers. Today, nobody would
build an HPC cloud for these particular big-
science grand-challenge applications. It simply
isn't a profitable business: the “HPC market” is
far too small and thus lacks economy of scale. In
some specific science application scenarios, with
complex workflows consisting of different tasks
(workflow nodes), a hybrid infrastructure might
make sense: cloud capacity resources combined
with HPC capability nodes, providing the best
of both worlds.
However, for a wide range of HPC applica-
tions like the parameter-sweeps mentioned above,
clouds will be the way to go. We already see more
and more HPC clouds today like Exa PowerFLOW
(Exa, 2008), and Cyclone (SGI, 2010) which offers
cloud services for engineering and scientific ap-
plications like BLAST, Gaussian, STAR-CCM+,
and LS-DYNA.

Grid-enabled applications require sustainable
Grid infrastructures. It doesn't make any sense,
for example, in a three-year funded Grid project,
to develop or port a complex application to a
Grid which will shut down after the project ends.
We have to make sure that we are able to build
sustainable Grid infrastructures which will last
for a long time. Therefore, in the following, the
author offers 'his' 10 rules for building a sus-
tainable Grid or cloud infrastructure, originally
presented in the OGF Thought Leadership Series
(2008). These rules are derived from mainly four
sources: research on major Grid projects published
in a RENCI report (Gentzsch, 2007a), the e-IRG
Workshop on “A Sustainable Grid Infrastructure
for Europe” (Gentzsch, 2007b), the 2nd Inter-
national Workshop on Campus and Community
Grids at OGF20 in Manchester (McGinnis, 2007),
and personal experience with coordinating the
German D-Grid Initiative (D-Grid, 2008). The 10
rules are mainly non-technical, because we believe
most of the challenges in building and operating
a Grid are in the form of mental, cultural, legal
and regulatory barriers. Although these rules have
been derived originally for successfully building
a sustainable Grid infrastructure, recent experi-
ence with cloud computing shows that most of
these rules still hold for introducing, building or
connecting to a cloud infrastructure.
Rule 1: Identify your specific benefits . Your
first thought should be about the benefits for your
users and your organization. What's in it for them?
Identify the benefits which fit best: transparent ac-
cess to and better utilization of resources; almost
infinite compute and storage capacity; flexibility,
adaptability and automation through dynamic and
concerted interoperation of networked resources,
in-house or from a public cloud; cost reduction
through utility model; shorter time-to-market be-
cause of more simulations at the same time on the