Features:
LEVERAGING LINUX CLUSTERS FOR OIL EXPLORATION
by Tim Curns, Editor
Founded in 1983, Weinman GeoScience provides expert exploration consulting and
seismic data processing services to the energy industry. Dedicated to the
effective integration of geological, geophysical, and engineering data,
Weinman Geoscience provides multidisciplinary solutions to the exploration
and exploitation problems of its clients.
Recently, HPCwire caught up with Dick Paschal, Senior Geophysicist at Weinman
Geoscience and Barry Weinman, President of Weinman Geoscience, to discuss the
company's use of Linux clusters for seismic data processing.
HPCwire: How did you become interested in Linux clusters? How are these
clusters leveraged to expand your business?
Weinman Geoscience: Linux clusters are quickly becoming a major technology in
seismic data processing and now almost every major exploration group boasts a
Linux cluster of dozens, sometimes thousands or processors. Linux clusters
offer many advantages that make them ideal for seismic processing. For
starters, seismic processing is naturally parallel, so clusters are a good fit
for this computing environment. Additionally, Linux clusters offer a
price/performance ratio that makes high performance computing available to
smaller companies, like Weinman Geoscience. With our Linux cluster, we are
able to apply compute-intensive processes to larger and larger data volumes.
HPCwire: What obstacles stood in your way before you switched to Linux
clusters? In what ways was your existing computing platform insufficient for
your endeavors?
WG: The computing power we had before was not nearly enough to complete the
size of jobs we needed to run. Our computing resources had to be fully
utilized in order to justify the cost of the equipment and we were hard
pressed to handle large projects without shutting down everything else on the
system. Since installing the cluster, we've realized the compute power on our
previous SUN platform equaled the power of one of our Linux Networx compute
nodes.
The cluster has enabled us to pass the benefits of high performance computing
directly to our customers. For example, before coming to us, one company had
drilled four consecutive dry holes at a cost of $5 million each. After coming
to Weinman Geoscience, we were able to process seismic data that led them to
hit eight for eight good wells-leading to the discovery of a half billion
barrels of oil. The cluster not only helped our customer find new hydrocarbon
reserves, but also made our computing price/performance more competitive with
larger contractors.
HPCwire: Please describe your process of procuring your first Linux cluster?
What are the clear pros and cons of switching to Linux clusters?
WG: We carefully researched different Linux cluster vendors that we had seen
successfully deliver cluster systems for seismic processing. We approached
four different cluster vendors and asked them to submit a bid based on our
specifications. After receiving the bids, we invited the vendors back to give
presentations on how their specific solution could meet our needs. After
reviewing all the bids and presentations, we were impressed with Linux
Networx's cluster expertise and sophisticated cluster management tools, so
they were chosen to build our first cluster.
Linux clusters offer many advantages, such as the high performance, cost
benefit over traditional supercomputers, and their scalability. But it's more
than the speed and cost of Linux clusters that makes them attractive. Linux
clusters bring a level of accuracy to our seismic processing that we would be
unable to achieve otherwise. Because of this, we are able to stay competitive
with major contractors and provide fast and accurate seismic data to our
customers.
On the flip side, Linux isn't as mature as Solaris for system administration
and Red Hat WS can be very expensive, and is still not as solid as Solaris.
HPCwire: What types of criteria were you looking for when deciding what
cluster solution to purchase?
WG: The main criteria we were looking for were price, reliability, performance
& manageability. Basically what this all boiled down to was computing
productivity. We needed a system that could process the most number of jobs
possible during its lifetime for the greatest return on investment possible.
This required the system to be stable enough where downtime was at a minimum.
Also, the system had to be easy to manage so we could focus on the results,
and not spend time managing the cluster. Of course performance plays a vital
role in productivity, so we needed a cluster that could produce results in a
timely manner.
HPCwire: What made you choose Linux Networx to provide your cluster? How did
they fulfill your criteria requirements?
WG: Linux Networx focuses its whole business on maximizing a customer's
computing productivity. Everything from the design of their systems to their
quick build-up time is all done so their customers can hit the ground running.
We were impressed with this approach and knew that a cluster from Linux
Networx would be a reliable workhorse that would help us achieve our computing
goals. Linux Networx also successfully demonstrated their expertise in Linux
clustering with their hardware design, cluster management tools, and training
programs.
HPCwire: How, specifically, is the cluster used? What kinds of things does the
cluster allow you to do? What applications are run with the cluster?
WG: The cluster is being used for compute-intensive applications and seismic
data processing. The cluster allows us to produce large data volumes of
seismic data, fast parameter testing, and fast turnaround for other compute
intensive applications (noise attenuation, multiple removal). Most
importantly, however, the Linux Networx cluster has leveled the playing field
so we are able to compete with larger companies.
Applications that run on the cluster include ProMAX Seismic Processing,
Paradigm Seismic Imaging, and Proprietary Seismic Imaging Software.
HPCwire: Some companies may be swayed by speed rather than productivity. So
how do vendors guarantee a highly productive cluster? In other words, what can
they do to ensure productivity?
WG: Whatever application you use, it's important that the Linux cluster vendor
you're working with is able to optimize your application to run on Linux
clusters. Without proper optimization, your application may not be able to
take advantage of the cluster system architecture. Look for a cluster vendor
who partners with many top-tier application vendors to optimize and pre-
integrate software onto systems.
Also, Linux clustering isn't a "one size fits all" endeavor and involves more
than connecting computers together. Clustering is about delivering a powerful,
comprehensive computing platform tailored and optimized to your organizationšs
specific needs. It's important to work with someone who is willing to make
sure the cluster system runs to your specifications.
HPCwire: What advice would you offer other oil & gas organizations looking to
purchase a Linux cluster?
WG: To find the best solution to meet your needs, you need to become an expert
on Linux clustering yourself. Research what different vendors offer and how
these vendors will help you maximize your computing productivity. Also, know
the requirements of your applications. Since the applications you use can
determine the design of the system, be familiar with these requirements and
design the system with the application(s) in mind.
HPCwire: Anything else you'd like to add?
WG: Just Do it!!!
|
