A number of Cray folks attended the Rice Oil & Gas workshop in Houston last Thursday, February 28th. It was incredible visiting such an influential group of energy leaders. Most of the major US oil & gas companies and service providers were there.
The morning keynote was presented by Dirk Smit, VP of Exploration Technology and Chief Scientist Geophysics, Innovation and R&D, at Shell. In the keynote, Compute Challenges to Meet Future Energy Demands, he stated that supplying the world’s water and food will become progressively critical. According to the LA Times, we’re adding more than 70-million people a year to the planet. Supplying water and food to them will demand new research and investments.
In upstream oil & gas exploration, a big challenge today is storing, processing, and interpreting seismic data. There is so much data, the big challenge is visualizing and transforming the data into useful information.
People often think of Cray in the context of supercomputers. What’s less known is that we also deliver massively scalable storage solutions for upstream oil & gas exploration. Workflows for data acquisition, processing, and interpretation of seismic and full wavefield inversion (FWI) create tremendous amounts of data. These workflows are well-suited to Cray’s storage offerings. The storage IO demands are growing such that parallel file systems and interconnects are being pushed to the max (often obviating the use of traditional SAN or NAS) for primary “close to compute” workflows.
In the future, according to Smit, data may need to move closer to compute. He used the analogy of data being like raisons in a compute cake. The data and IO simply needs to be there. Available and accessible when needed—without the latency of retrieval. Today IO is separated by networks—IP, fiber channel, InfiniBand, for example. Smit questions the viability of current network-attached models for storage IO. Will they be fast enough to handle the upstream exploration workloads of the future?
At Cray we consider these storage and big data challenges in the context of Excascale. Storing and protecting really big data sets—as immediately accessible—requires new architectures to dispatch the right storage IO depending on the application. Today we talk about “Big IO”. Fast forward to tomorrow’s architectures, where Cray systems will dispatch data to compute in real-time based on application or code need. One building block technology for achieving near-real-time IO across heterogeneous file systems is Cray Data Virtualization Services.
One of the most entertaining and informative speakers of the day was John Kuzan, Manager for ExxonMobil’s Upstream Research business. In Kuzan’s afternoon keynote, Kuzan said he expects performance—and that his HPC systems are accessible and adaptable. Kuzan does not want to have to call an HPC specialist to get work done. Years ago, proprietary systems often required a specialist to assist. Today, Kuzan can remote in from his PDA, update models on the fly, and his team can do more on their own.
Another speaker was Bill Kramer, Deputy Project Director for the NCSA Blue Waters project at the University of Illinois. In Kramer’s presentation, he discussed how the NCSA Blue Waters project uses Cray for compute storage. The open Cray Linux Environment (CLE) uses Fortran, C, C++, Python, and UPC for a broad range of codes and applications. The platform is completely open and adaptable—supporting a broad range of compilers, debuggers, IO libraries, tools, and workloads. For storage, Cray packages these data management tools externally in a development platform (esLogin).
At Blue Waters, they’re using Cray Sonexion to store over 26 petabytes of usable storage. 23 petabytes are in a single Lustre file system. This is online storage, with aspects converging on nearline. Keeping the data on disk is also important for accessibility. Although the systems supports a massive tape archive, users prefer not to wait for data to be retrieved from tape.
Kuzan said ExxonMobil constantly evolves their upstream research business. It’s incredible how far Exxon has come and were they’re going.
Some of this evolution is tracked in their corporate magazine, The Lamp. An interesting factoid about The Lamp, according to Kuzan, is the name: The Lamp was named in honor of the switch from whale oil to petroleum for fueling household lighting lamps in the mid-1800’s. It’s a good read. Check out ExxonMobil’s latest edition of The Lamp. Cray is mentioned in the context of seismic and FWI.
Jason Goodman, Cray Storage & Data Management