Supercomputing for oil and gas is heating up, as giant energy firms race to exploit the advantages of HPC for their seismic imaging workloads. Energy goliath BP recently revealed a new datacenter in Houston that will house what BP is calling the “world’s largest supercomputer for commercial research,” clocking in at more than 2.2 petaflops. The system and its brand-new digs are part of a five-year, $100 million HPC infrastructure investment program at BP.
Back in March, French oil and gas rival Total Group announced it would be installing the “Pangea” SGI ICE-X cluster at the company’s Scientific and Technical Centre in southwest France. Pangea packs 110,400 cores and 7 PB of storage capacity and runs at 2.3 petaflops, casting a shadow on BP’s claim of owning the largest civilian supercomputer. The contract with SGI, which is valued at $77.3 million, calls for Pangea’s performance to be doubled in 2015.
The latest oil and gas company to join the HPC club is Woodside Energy. According to an article at IT News, Woodside, Australia’s largest independent oil and gas firm, is putting an initial $500k towards a Xeon-powered supercomputer to support its oil exploration efforts.
The initial system will provide 0.118 petaflops of processing power by the end of the year, but the company hopes to scale up to petascale territory by 2015. A Woodside spokesperson confirmed to IT News that the machine is an Intel system with 110 Xeon E5-2670 processors, 55 nodes and 880 cores total. Unlike BP or Total, which are building straight-up CPU-based machines, Woodside plans to add accelerator hardware – most likely NVIDIA Tesla GPUs – to each of the 55 nodes.
The supercomputer has been named Moordiup – the Aboriginal word for “fast.”
Woodside’s spokesperson told IT News that Moordiup will enable the company to do “advanced seismic processing” in-house, where previously they contracted with external service providers, who did mainstream seismic processing. A key first task will be tackling a geophysics problem called full wave inversion (FWI), which uses normally discarded 3D seismic data to generate more accurate models of oil and gas reserves.
“The idea is for us to test FWI on a small scale and also to test the hardware with the intention that if we can get all these things to work we will expand Moordiup to a Petaflop-scale machine which we think we’ll need for a production environment by about 2015,” said Woodside subsurface technology chief scientist, Tom Ridsdill-Smith.