HPC solutions and the oil and gas sector
Seismic imaging and reservoir simulation are the most compute intensive tasks performed in the oil and gas industry. Improvements in seismic image clarity, resolution, and accuracy have direct financial impact financial in a business where one exploration well in deep water will cost in excess of $100M. While there are no known seismic hydrocarbon indicators, a company that gets the complete three-dimensional image correct has a better chance of finding productive hydrocarbon traps. Then, once production is established, the development plan for a field is entirely dependent on relating a well’s performance to the structure and internal characteristics of the producing zones. Therefore improving seismic detail is vital to maximizing hydrocarbon recovery. An interpretation based on poor data can have serious business implications. Oil and gas companies can buy a strategic edge by managing data-intensive tasks faster and better than the competition.

As seismic imaging applications grow in complexity the algorithms become constrained by data movement and scalability challenged. In addition, as the industry moves towards next generation imaging algorithms such as 3D Full Waveform Inversion (FWI) which promise much greater image fidelity, there is an even greater emphasis placed on the computing and storage infrastructure.

Cray and VRBX are facilitating the discussion
The challenge is that finding success in this area is not just a matter of building the biggest supercomputer in the industry and winning the HPC arms race.    Instead, oil and gas companies have to combine the right HPC architectures for their needs with advanced applications that support data-intensive research and production. This is often a complex and time-consuming process.   The workshop aims to help oil and gas companies identify the best ways to use advanced HPC technologies and applications by discussing industry-leading upstream oil and gas applications – such as FWI -, ways to accelerate application functionality, challenges that arise when developing HPC infrastructure and roadmaps for HPC system and storage deployment.

You can view a full agenda here http://www.cray.com/events/oil-gas-workshop-brazil.html

Per Nyberg, Director of Business Development 

Chapel originated from the DARPA High Productivity Computing Systems (HPCS) program, which challenged vendors like Cray to improve the productivity of high-end computing systems.  Engineers at Cray noted that the HPC community was hungry for alternative parallel programming languages and developed Chapel as part of our response.  The reaction from HPC users so far has been very encouraging—most would be excited to have the opportunity to use Chapel once it becomes production-grade.

Chapel Overview 

Though it would be impossible to give a thorough introduction to Chapel in the space of this article, the following characterizations of the language should serve to give an idea of what we are pursuing:

• General Parallelism: Chapel has the goal of supporting any parallel algorithm you can conceive of on any parallel hardware you want to target.  In particular, you should never hit a point where you think “Well, that was fun while it lasted, but now that I want to do x, I’d better go back to MPI.”
• Separation of Parallelism and Locality: Chapel supports distinct concepts for describing parallelism (“These things should run concurrently”) from locality (“This should be placed here; that should be placed over there”).  This is in sharp contrast to conventional approaches that either conflate the two concepts or ignore locality altogether.
• Multiresolution Design: Chapel is designed to support programming at higher or lower levels, as required by the programmer.   Moreover, higher-level features—like data distributions or parallel loop schedules—may be specified by advanced programmers within the language.
• Productivity Features: In addition to all of its features designed for supercomputers, Chapel also includes a number of sequential language features designed for productive programming.  Examples include type inference, iterator functions, object-oriented programming, and a rich set of array types.  The result combines productivity features as in Python™, Matlab®, or Java™ software with optimization opportunities as in Fortran or C.

Chapel’s implementation is also worth characterizing:

• Open Source: Since its outset, Chapel has been developed in an open-source manner, with collaboration from academics, computing labs, and industry.  Chapel is released under a BSD license in order to minimize barriers to its use.
• Portable: While Cray machines are an obvious target for Chapel, the language was designed to be very portable.  Today, Chapel runs on virtually any architecture supporting a C compiler, UNIX-like environment, POSIX threads, and MPI or UDP.
• Optimized for Crays: Though designed for portability, the Chapel implementation has also been optimized to take advantage of Cray-specific features.

Chapel: Today and Tomorrow 

While the HPCS project that spawned Chapel concluded successfully at the end of 2012, the Chapel project remains active and ongoing.  The Chapel prototype and demonstrations developed under HPCS were considered compelling enough to users that Cray plans to continue the project over the next several years.  Current priorities include:

• Performance Optimizations: To date, the implementation effort has focused primarily on correctness over performance.  Improving performance is typically considered the number one priority for growing the Chapel community.
• Support for Accelerators: Emerging compute nodes are increasingly likely to contain accelerators like GPUs or Intel® MIC chips.  We are currently working on extending our locality abstractions to better handle such architectures.
• Interoperability: Beefing up Chapel’s current interoperability features is a priority, to permit users to reuse existing libraries or gradually transition applications to Chapel.
• Feature Improvements: Having completed HPCS, we now have the opportunity to go back and refine features that have not received sufficient attention to date.  In many cases, these improvements have been motivated by feedback from early users.
• Outreach and Evangelism: While improving Chapel, we are seeking out ways to grow Chapel’s user base, particularly outside of the traditional HPC sphere.
• Research Efforts: In addition to hardening the implementation, a number of interesting research directions remain for Chapel, including resilience mechanisms, applicability to “big data” computations, energy-aware computing, and support for domain specific languages.

For more information

For more information about Chapel, the best introduction is A Brief Overview of Chapel, which provides a concise summary of Chapel’s history, motivating themes, and features.  Further information, such as tutorials, presentations, and papers, can be found on the project website at http://chapel.cray.com.  To download Chapel or join various community mailing lists, visit our SourceForge page at:  https://sourceforge.net/projects/chapel/.

Brad Chamberlain, Principal Engineer 

A close look at the XC30-AC
On the inside, the Cray XC30-AC supercomputer is not all that different from our liquid-cooled Cray XC30 system. The AC-version leverages the same HPC integration foundation, compute blade, processors, accelerators, daughter cards, coprocessors and network systems as the already-popular liquid-cooled Cray XC30 system. The difference between the systems comes from a couple sources – how they are cooled, powered and how they are packaged.

The liquid-cooled Cray XC30 supercomputer uses a horizontal blade setup to and a transverse blower cabinet configuration to send air through the HPC system while liquid coil panels provide cooling in each compute cabinet. Alternatively, the Cray XC30-AC system utilizes a highly efficient bottom-top fan architecture that we validated in the successful XE/XK generation of Cray supercomputers. The Cray XC30-AC system is also unique in that it is deployed in smaller cabinets that are racked vertically and not as densely packed as its liquid-cooled counterpart. The Cray XC30-AC supercomputer offers both 208V and 480V power options and can be built in configurations ranging from one to eight cabinets.

Usage models for the Cray XC30-AC
Industries like, energy, finance, engineering and manufacturing are increasingly dependent on advanced simulations to support research efforts. High-performance computing solutions are integral in these settings, but many organizations in the technical enterprise market space cannot afford to implement the capability supercomputing systems needed to support such functions. Furthermore, many supercomputers use large quantities of power and require specialized data center architectures to support them. The Cray XC30-AC system provides integration, scalability and adaptive programming capabilities that make the system an ideal fit for companies that want to run advanced simulations, but on more frugal budgets than those needed to support a traditional top-end HPC solution.

Cray and technical enterprise
The Cray XC30-AC supercomputer is not our first foray into the technical enterprise. As one of the world’s leaders in supercomputing, we have  had  success with previous generations of technical enterprise offerings, and see increasing demand from users and influencers – engineers, scientists, researchers and business executives. The Cray XC30-AC system is, to a great extent, a result of that knowledge and experience. Cray does not just value the major labs. Instead, we understand the importance that HPC solutions can play in the technical enterprise space and have developed a solution to meet those unique requirements.

Jay Gould, HPCS Product Marketing Manager 

CUG is an independent corporation of member organizations that own Cray systems. It actually runs independently of Cray Inc., but with very close ties.  Cray employees serve as Special Interest Group (SIG) leaders, submit and deliver papers at the conference and actively participate with other CUG conference attendees. CUG is an invaluable way for the leadership at Cray to collaborate with our customers, celebrate their achievements, listen to their challenges and ideas, and get feedback on our future plans.

Personally, I have had the pleasure of serving as the Cray liaison to the CUG board of directors for the past several years. I’ve been able to work with them to ensure that Cray supports the organization however we can and that we are available to our customers, not only at the annual meeting, but throughout the year.

I’m looking forward to listening to the talk by Horst Simon on “Why we need Exascale, and why we won’t get there by 2020.” And as with my fellow Cray attendees, I’m always curious about what happens behind the closed doors of the “Cray 1:100 or more” session with our CEO, Peter Ungaro. This tradition, which was started about five years ago, gives our customers the opportunity to talk openly with Pete without any Cray employees or partners in the room. It would be interesting to be a “fly on the wall” for that one.

And it looks like CUG has a fabulous culminating event planned for Wednesday night at the Castello di Amorosa, a winery in the valley that was inspired by the medieval castles of Europe. It even has its own dungeon —complete with torture devices. Hmm… wonder if they have something planned for the Cray executives.

While not a wine connoisseur myself, I’m still looking forward to my upcoming trip to Napa. I always come home from CUG with a great sense of pride. As a Cray employee, it’s easy to focus on the products and services we provide, but seeing and hearing how Cray systems are making a difference in the real world is what I really get excited about.

Christy Adkinson, Director of Corporate Marketing

HPC systems and climate research
Climate studies often use similar types of simulations to the ones used in weather forecasting, but with an emphasis on the interactions and feedbacks in the entire earth’s system.  HPC solutions have enabled meteorologists to, for example, simulate how weather patterns will move over the course of several hours or days and predict the most likely outcomes with great precision. Climate models simulate the earth’s system for seasonal, decadal and centennial projections.   This allows researchers to identify long-term weather patterns, predict the likelihood of extreme storms and help government organizations prepare for climate change.

The complexity of this research is clear in the recent UPSCALE project. Led by the U.K. Met Office, the National Centre for Atmospheric Science (NCAS) and the University of Reading, UPSCALE aims to take the high resolution weather models used to predict conditions for six-day periods and apply them to 25 years. This type of simulation could offer incredible insight into future climate patterns, but also required HPC capabilities that were out of the reach of the U.K.-based organizations. However, the group heading the research looked to the Partnership for Advanced Computing in Europe, which gave the UPSCALE group access to the Cray XE6 supercomputer nicknamed “Hermit” at HLRS in Stuttgart, Germany, making the advanced high-resolution simulations possible.

Cray and climate research
Cray is among the leading supercomputing vendors when it comes to supporting climate research. Understanding how the global climate is changing and what organizations can do to prepare is critical to humanity’s future and Cray is playing a leadership role in helping governments and research centers understand extreme weather and climate patterns. Besides the UPSCALE project, Cray is also enabling extreme weather event studies and hurricane modeling efforts, among other prominent projects.

Jay Gould, HPCS Product Marketing Manager 

Making strides in genetic research
The project, which was recently detailed in the Genes, Brain and Behavior journal, used data-intensive supercomputing to complete sophisticated research that would have been impossible without such technology. A team of scientists from the San Diego Supercomputing Center (SDSC) and the Institute Pasteur have identified a time-dependent gene-expression process that could help medical professionals treat mental disorders such as schizophrenia and autism.

Within the hierarchical tree of various coherent gene groups, there are transcription-factor networks in which a variety of genetic patterns develop. These patterns typically form during the brain’s development and often dictate various aspects of a person’s mental makeup. As a result, the master transcription factors that are at the top of coherent gene groups often hold the key to understanding how autism, schizophrenia and similar disorders develop.

In the National Science Foundation press release detailing the research, Igor Tsigelny, a research scientist with SDSC and UC San Diego’s Moores Cancer Center, explained that data is central to these types of advances.

“We live in the unique time when huge amounts of data related to genes, DNA, RNA, proteins, and other biological objects have been extracted and stored,” said Tsigelny.

Using very large genetic computing data simulation to make meaningful progress
Gathering and using genetic data can prove extremely challenging. For this genetic research project, progress was made through the use of SDSC’s Gordon supercomputer, which is a Cray CS300-AC™ Cluster Supercomputer. Gordon’s I/O nodes are specifically designed to handle large, complex data-intensive workloads that address I/O bottlenecks. Researchers and scientists at the SDSC facility have been using the Gordon supercomputer since January 2012, and the system is now part of NSF’s Extreme Science and Engineering Discovery Environment (XSEDE) program — a nationwide partnership including 16 high-performance computers and high-end visualization and data analysis resources.

Gordon’s robust data I/O capabilities have been used by many scientists in a variety of fields, whose research requires the mining, searching and/or creating of large databases for immediate or later use, including mapping genomes for applications in personalized medicine and examining computer automation of stock trading by investment firms on Wall Street.
For more details on how the Gordon supercomputer has impacted genetic research, check out this NSF video.

Maria McLaughlin, Cray Cluster Solutions, Marketing Director 

Friedlander’s 79 framed photographs are presented in a circular format in one room, each frame pressed against the next in a story that slowly circles from the countryside of Chippewa Falls, Wis., into the town and finally to the people that have shaped the pinnacle of computing technology for nearly half a century.

In this small space and format, the photographs tell a story as unique and powerful as any of the legendary stories of 20th-century technological innovation. Like Hewlett and Packard in their Bay Area garage or Jobs and Wozniak in theirs, Seymour Cray and the legacy of this small Wisconsin town still reverberate through the lives of scientists and engineers around the world as well as through the lives and families of this community. It reminds us that our community of technologists is bound to communities like Chippewa Falls and each shapes the other. Imagining these same people – captured working with such focus in Friedlander’s photographs – as merchants, farmers or timber workers just 20 years earlier and then leaping forward to today and realizing that their children and grandchildren are still producing these magnificent machines that shape our understanding of the world and the universe is sobering and exciting.

Today Cray is a company with many communities at its core. The cultures and attitudes of Minnesota, Wisconsin, Washington state, Silicon Valley, South Korea, Texas, Japan and the countries of a unified but diverse Europe shape Cray today in a concrete and fundamental way. Lee Friedlander’s photographs remind us that Cray and all of high performance computing is shaped through the families and communities from which we have evolved. To some degree, everyone in supercomputing today is a “Crayon” and whether we know it or not, we are all Chippewaians, owing a large part of who we are as an industry to that small community.

The exhibit runs through June 16^th and if you’re in the Bay Area, it’s certainly worth a stop.

Barry Bolding, Vice President of Storage & Data Management and Corporate Marketing

Lustre has evolved and matured into the pre-eminent, open parallel file system for HPC and big data storage. Our open storage system strategy builds on the Linux® operating system (across numerous distributions), and combines interoperability, platform choice, and the Lustre file system.

Why Lustre?  Simply put: performance, scalability, and innovation through open collaboration.

The Lustre file system powers around 2/3rds of the top 100 HPC sites.  At Cray, we have a broad number of Lustre deployments including our Cray Lustre file system optimized for NetApp and DDN storage, and Lustre is also fully embedded in our Cray Sonexion™ storage system – our scale-out Lustre system for HPC and big data.  You can read more about Cray’s Lustre expertise for Big Data and HPC in this solution brief.

There has been some mention of Lustre achieving 1TB/s to a single file system.  Through the evolution and maturity of Lustre, and support from partner companies, we have worked hard to  deploy and achieve sustained aggregate throughput of 1TB/s to a single (23 petabyte) file system for the NCSA Blue Waters project.   Everyone at Cray is extremely proud of this accomplishment, and honored to provide Blue Waters users with application-centric computing and storage built on Lustre.

According to project leader Bill Kramer, in our case study on Blue Waters, Enabling Scientific Breakthroughs at the Petascale, Cray’s implementation of Lustre provides high levels of scalability, reliability and stability for continuous operations in an open-source, non-proprietary environment. “I’m not sure we could scale any other file system at this point, and even if it could be done with another approach it is even more unlikely we could have done it for the same cost profile,” Kramer said.

The open source community struggled early on to solidify Lustre, but the quality and maturity of the Lustre implementations and source code base has vastly improved and stabilized over the past few years to enable petascale deployments like Blue Waters.  Major companies, including Cray, have invested heavily in development to enable the  use of Lustre, formed close partnerships, and validated Lustre as a reliable and stable platform for major production application workloads, within the national labs and out across commercial industries.

This has been achieved through a collaborative consortium called Open Scalable File Systems (OpenSFS).  Cray, DDN, Lawrence Livermore National Laboratory and Oak Ridge National Laboratory founded OpenSFS in 2010. The founding members have since been joined by Xyratex Corp. and Intel, Inc. at the Board Level, plus the Board is completed by a Community Representative Director, a position elected yearly. The CRD for the 2013 term is from the Texas Advanced Computing Center.

One of the great events that bring these partner organizations  together is the Lustre User Group (LUG).  This year, LUG is  being held in San Diego, and it’s sure to be filled with good technical papers and informative presentations.  Check out the OpenSFS LUG 2013 website for more information. https://www.gaveledge.com/SFS1301/agenda.

You can also read our latest white paper  ‘Cray Brings Leadership and Customer Focus to Lustre Scalability’.

Jason Goodman, Cray Storage & Data Management

YarcData Blog: Reflecting on our Graph Analytics Challenge

By Arvind Parthasarathi, President

I wanted to thank the participants of our $100,000 Graph Analytics Challenge. It took our judges much deliberation to choose the winner, because each of the six finalists presented a complex problem and offered a unique, innovative solution. The entries spanned a number of diverse topics, from medical research to social collaboration to sports analytics. I could not be more pleased that graph analytics is drawing such high calibre experts who are so passionate about their work to better society.

As graph analytics gains traction, customers and analysts have asked me how I believe graph analytics can improve existing technologies and address real business concerns. Apart from my opinion, our contest has demonstrated the applicability of graph analytics to high impact use cases. From discovering a cure for autism or preventing crimes to predicting baseball outcomes – all are important issues with significant business and human impact.

The Graph Analytics Challenge has shown that the most complex problems involve discovering the previously unknown. Discovery is challenging because you don’t know in advance what queries you will run, or what data you will need. To quote Ilya Shmulevich from the Institute of Systems Biology: In the amount of time it takes to explore one hypothesis, we can now explore thousands of hypotheses, massively improving our success rate, I think that summarizes the YarcData value proposition in a nutshell.

I’d like to extend my thanks to all our contest participants and especially to our first, second, and third place winners respectively: Ilya Shmulevich, Brady Bernard, and Andrea Eakin, of The Institute for Systems Biology; Adam Lugowski, John Gilbert, and Kevin Dewesse, of the University of California at Santa Barbara; and Abraham Flaxman of the University of Washington Institute for Health Metrics and Evaluation. We look forward to seeing where all your research takes us into the future and how through systematized Discovery you are making the world a better place.

To learn more about the contest and other Big Data discussions check out the YarcData blog. www.yarcdata.com/blog

Cray’s presence at the Bio-IT World Conference
The life sciences community is becoming more dependent on data-intensive research as a means to achieve knowledge discovery gains. As life sciences continue to evolve, the best way to support knowledge discovery increasingly depends on being able to process and analyze extremely large quantities of data. Carlos P. Sosa, Manager of the Chemistry and Life Sciences Segment at Cray, will present on how Cray’s high-performance computing systems are contributing to major knowledge discovery advances in life sciences.

Sosa’s presentation – Building Bridges: Evolving High Performance Computing in the Life Sciences – will take place at 2:15pm on Wednesday, April 10. He will focus on how Cray is engaging the scientific community to develop better high performance computing solutions that enable enhanced research and develop next-generation applications.

An example in aligning science and computing will also be presented at the 2013 Bio-IT World Conference when Jack Wells, Director of Science, Oak Ridge Leadership Computing, presents about the research being done using the Cray Titan supercomputer.

Considering Cray’s impact on life sciences
Cray has had a major impact on the life sciences community by helping researchers complete important projects. A few of these include efforts from research groups at the University of Chicago through the use of the Beagle supercomputing system, such as:

• Taeyoon Kim’s group at the University of Chicago – Using computational modeling to support research exploring how a cell contracts with an external matrix.
• Aaron Dinner’s group at the University of Chicago – Testing new software with the help of high-performance computing systems.
• Rick Stevens’ group at the University of Chicago and Argonne National Laboratory – Built flux balance analysis (FBA) based metabolic models for all sequenced microbes using high-performance computing.
• Maryellen Giger’s group at the University of Chicago – Has used high-performance computers for advanced breast cancer image analysis and computer aided diagnosis.

To learn more about Cray’s work in the life sciences community stop by booth 334 at the Bio-it conference.