When it comes to oil and gas supercomputers, the ante has been upped once again. In the past couple years, energy giants BP, Total and Eni have rolled out mega supercomputing centers, and now Norwegian oil and gas company Petroleum Geo-Services (PGS) has outdone them all with a 5-petaflops Cray XC40 supercomputer and a Cray Sonexion 2000 storage system.
Headquartered in Oslo, Norway, with sites in London, Singapore and Houston, PGS depends on supercomputing to create high-resolution seismic maps and 3-dimensional (3D) models of the earth’s subsurface. Earlier this week, Cray’s Barry Bolding filled in some details about the big win, which is the largest of the publicly-announced oil and gas petafloppers.
A key company in the oil and gas exploration space, PGS navigates the globe, gathering sensor data by bouncing waves off the substructure of the earth’s crust. Marine exploration is a speciality and one of their biggest endeavors is the PGS Triton survey, a sophisticated seismic imaging survey being conducted in the deep waters of the Gulf of Mexico.
A long-time consumer of computational and storage products, PGS like other energy companies, is seeing large growth in the data sets that they are acquiring and there is a consequential need for a more performant simulation environment to process the data and do so at high resolution.
“With access to the greater compute efficiency and reliability of the Cray system, we can extract the full potential of our complex GeoStreamer imaging technologies, such as SWIM and CWI,” said Guillaume Cambois, executive vice president Imaging & Engineering for PGS in an official statement.
From Bolding’s perspective, Cray was able to provide the right combination of high performance supercomputer and high performance storage for PGS’s needs. The XC40 supercomputer has the right mix of components for the kind of computationally-intensive big physics math problems used in seismic processing. “The high performance interconnect, high-performance scalability, scalable software, compilers and libraries are just the infrastructure PGS needs to be able to scale their problems up and get the high resolution imaging they require,” stated Bolding.
Fast storage is also required in order to move the large datasets in and out without bottlenecks. According to Bolding, the Cray Sonexion Lustre storage system fit the customer’s need for a fast parallel file system.
Although the Cray XC40 was designed to work with both NVIDIA GPU Tesla accelerators and Intel’s Xeon Phi coprocessors, PGS opted for a standard x86 computing environment with Intel “Haswell” processors for the system, which sits right in their production workflow. For the record, BP and Total are also running CPU-only machines, while the Italian energy company Eni is running its workloads on a hybrid CPU-GPU cluster architecture with 1,500 dual CPU nodes and 3,000 NVIDIA GPUs.
Although Cray is not at liberty to discuss many of its commercial customers, Bolding shared that energy is an important sector for the company and accordingly Cray has sales and pre-sales staff dedicated to these customers.
Bolding also talked about how analytics is becoming more important in the energy sector. He says there are opportunities for analytics in the oil and gas space for downstream analysis, but even more interestingly is the potential role for analytics methodologies in the upstream portion of energy exploration, which is a new area for both Cray and for customers, and one where experimentation is just beginning.
In closing, Bolding says he sees this contract as an example of real growth in the energy segment, especially where the datasets are becoming very large.
The systems are scheduled for deployment later this year. The dollar value of the sale was not disclosed by either party, but other XC40 contracts have ranged from $9 million all the way up to $128 million – and the “Trinity” deal that Cray inked with the National Nuclear Security Administration is valued at more than $174 million.