Cray, SGI Ink Supercomputer Deals With Energy Companies

Cray will bring a 5-petaflop system to oil-and-gas company PGS, while SGI will upgrade Total's ICE X Pangea supercomputer.

Cray supercomputer

Cray and SGI have been pulling in significant supercomputer deals as commercial organizations look to leverage petaflop capabilities for their businesses.

Cray will build a 5-petaflop supercomputer for Petroleum Geo-Services (PGS), a global oil-and-gas company based in Oslo, Norway, that develops high-resolution seismic maps and 3D models of the sub-surface of the Earth that other businesses in the field use as they look for more energy resources.

For its part, SGI officials announced March 31 that the company will upgrade the SGI ICE X high-performance computing (HPC) system that currently is in use by another energy company, Total, whose work involves everything from finding and producing oil and natural gas to chemicals, marketing and new energy sources. At the same time, SGI Japan—part of SGI—is building an ICE X server cluster that will be deployed as part of a supercomputer being used by Japan's National Institute for Materials Science (NIMS).

For Cray, the deal with PGS is part of a larger effort to rapidly grow the business not only around supercomputers, but also storage and analytics systems. During a conference call in February to talk about the latest quarterly financial numbers, Cray CEO Peter Ungaro said the beginning of 2015 was promising.

"On the heels of a great year of new wins in 2014, we came into 2015 on a good pace to achieve our goals, as we have more current year business already in place than we'd ever had at this point, both in dollars and as a percentage of our full year outlook," Ungaro said, according to a transcript on Seeking Alpha. "While we still have work left to do to secure new wins for 2015, we're in really good shape at this point."

The systems vendor didn't release financial details related to the PGS deal, but said it involves both a supercomputer and a storage system. Cray will supply an XC40 supercomputer and Sonexion 2000 storage system, with the supercomputer being one of the largest the company has ever delivered to a commercial customer.

The systems, which will be deployed later this year, will give PGS the computation capabilities it needs to run its complex seismic processing and imaging capabilities, according to Cray officials. The 5-petaflop XC40 includes Cray's Aries interconnect technology, its Dragonfly network topology for creating low-latency networks and the option of the DataWarp solid-state drive (SSD) accelerator technology. The XC40, which was launched in September 2014, also includes a variety of processor types, including Intel's x86 Xeon processors and Xeon Phi co-processors and Nvidia's Tesla GPU accelerators.

The supercomputer also offers Cray's Linux environment and its HPC-optimized programming technology.

SGI will add another 4.4 petaflops of performance to Total's current Pangea 2.3-petaflop system, according to company officials. The vendor will add a new ICE X system that includes the company's M-Cell power and cooling technology, storage and Intel's latest Xeon E5-2600 v3 processors. The performance would put the supercomputer into the top 10 on the most recent Top500 list of the most powerful systems, which was released in November 2014, SGI officials said.

"Oil and gas companies have faced increasing difficulty in the discovery and extraction of new oil and gas reserves in recent years, making HPC technology a crucial tool for organizations in this industry," SGI President and CEO Jorge Titinger said in a statement.

Total's current Pangea system is powered by Xeon E5-2670 v3 chips and holds 110,592 cores. It also contains 442 terabytes of memory, and includes SGI's InfiniteStorage 17000 disk arrays, Intel's Enterprise Edition for Lustre File system and SGI's DMF tiered storage virtualization. The updated system will include another 9.2 petabytes of storage, 4,068 more nodes based on Xeon E5-2680 v3 chips that comprise 110,592 cores, and 589 terabytes of memory.

In Japan, the new ICE X system was deployed in December 2014, though the company announced it this month. It's a scale-out platform with 1,044 nodes, the latest Xeon E5-2600 v3 chips and 25,056 cores. It offers 130TB of memory and a performance of 1 petaflop. It offers a 1.4-times density improvement over the previous system, SGI officials said.

It's connected to 675TB of shared physical storage via an InfiniBand interconnect and uses a Lustre file system.