Special Features:
SHELL E&P USES LINUX NETWORX GRID CLUSTER TO
BOOST RESEARCH
Overview
Shell International
Exploration & Production (Shell E&P) needed a
supercomputing solution to
facilitate development and deployment of new
algorithms in the areas of
seismology, geomechanics, reservoir simulation, and
large-scale inverse
problems.
The primary class of research projects being addressed at
Shell E&P required
large memory and high-speed internode communication. With
their previous
supercomputing systems, Shell E&P scientists were limited in
the types of
algorithms they could effectively develop and deploy.
Linux NetworX provided Shell E&P with three grid subclusters totaling 112
processors configured with hardware to boost current research and development
efforts. With the grid cluster from Linux NetworX, Shell E&P scientists are
able to expand the scope of their research and the types of problems
addressed.
Challenge
Shell E&P is actively involved in many
research projects in the areas of
seismology, facture mechanics, and
multi-phase reservoir simulation. With a
long history in high-performance
computing, Shell E&P began using Cray
machines in the early 1980s and
continued with many other platforms including
a variety of clusters,
workstations, and used external supercomputing
facilities to aid with their
research endeavors.
Although powerful, the systems Shell E&P were
using were not cost-effective
and did not have the level of scalability
required for the types of algorithms
scientists needed to advance their
research.
"Our thinking was restricted by the hardware we were using,
which limited the
types of problems we could practically hope to develop and
deploy," said Jim
Clippard, senior research geophysicist at Shell
International E&P.
"We needed to explore cost-effective, scalable
computing options with faster
internode communication and larger memories to
address many problems that we
were ignoring due to intrinsic limitations or
cost of our existing hardware."
Solution
To expand their
research capabilities, Shell sought out Linux NetworX, a
producer of Linux
grid cluster supercomputers, to build and configure a
cluster to maximize
research and development capabilities. From previous
experience, Shell
researchers knew that different algorithms benefit from
different hardware
configurations. To overcome this problem, Linux NetworX
built three grid
subclusters totaling 112 processors with different
configurations for each
algorithm.
"We have been able to implement and test some algorithms
that would not have
been practical without the Linux NetworX grid cluster,"
said Clippard. "The
capabilities that the grid cluster is providing are being
heavily used and
some results of these efforts are already finding their way
into mainstream
processing."
Results
Since implementing the
grid cluster, Shell scientists have been able to
further expand the scale and
scope of their exploration and production
research endeavors and the types of
problems they address.
With unique hardware requirements for their
112-processor cluster, Shell was
impressed with the clustering expertise of
Linux NetworX.
"Linux NetworX has a high level of technical expertise
with respect to large
bandwidth internode communication, large memory
configuration, kernal tuning,
high-performance hardware, and the Linux Open
Source operating system in
general," said Clippard. "Linux NetworX confidently
promised to do whatever
was required to deliver an optimal integrated system
to our premises and in
accordance with our specifications. This was a
substantial commitment in light
of our unusual configuration. Linux Networx
ultimately delivered in accordance
with our specifications and we have
realized the performance gains that we
anticipated."
|
