The latest GPU hybrid system, HPC4, that Eni has used to expand the computing capacity of their Green Data Center was supplied by HPE. Packed with an enormous 3,200 Tesla GPU accelerators, it is expected to modernize the process of oil exploration and gas activity management.
The procedure of finding and producing hydrocarbons is one of the most complicated & challenging ones. There are quite a few established techniques for discovering new hydrocarbon reserves. Seismic imaging of the subsurface is the basic one among those techniques. The proprietary implementation of anisotropic reverse time migration (RTM) is also quite popular these days. Eni has developed various advanced and fully integrated applications for geological and geophysical studies over the past 15 years.
In the RTM process, measurements are recorded at the surface by an array of receivers. A very accurate image of the subsurface of the Earth is then formed using the RTM algorithm. It takes 3D cubic grids extending over hundreds of square kilometers, down to 10 to 15 kilometers in depth to perform the computations that give the same size 3D image of the subsurface. It contains several billions of pixels and usually comes with a typical resolution of the order of 10 to 25 meters. Geologists and geophysicists use it to look for hydrocarbons.
The Anisotropic RTM technology is very useful for exploring substances lying beneath the Earth’s surface at places where the geology is quite complex. Up until recently, the anisotropic RTM was applied mainly in subsalt environments or while performing deep-water explorations. Its performance while processing the low-frequency seismic data is quite exceptional. One of the drawbacks of this method is that the computational expense and timing of running RTM on traditional compute clusters do affect the image resolution and the accuracy of results.
The latest GPU-accelerated computing has enabled Eni to fasten the anisotropic RTM process 4-5 times, making it possible to use the RTM process regularly in a wide variety of complex geological contexts. The results show more clarity, accuracy, and detail, as expected. With the fast, high-resolution seismic imaging and geological data in a single HPC platform, Eni is giving robust geological models that can revolutionize the exploration process of new hydrocarbon resources and reservoir management.
The HPC4 supercomputer of Eni resides at Eni’s Green Data Center in Ferrera Erbognone, near Milan. As of now, the HPC4 has turned its HPC infrastructure into the world’s most powerful industrial computing system. The 3,200 NVIDIA Tesla P100 GPU accelerators in collaboration with the NVIDIA Tesla K80 GPUs, have pushed Eni’s computational peak capacity to 22.4 petaflops.
“We started our path of supercomputing using hybrid clusters in 2013. We were the first in the O&G industry, and in 2014 we were awarded the HPCwire prize for the best use of HPC in the oil and gas sector. – Luca Bertelli, chief exploration officer at Eni
Eni also added “All our HPC systems from the beginning till now have been engineered with the same philosophy of hybrid architecture: powerful and energy-efficient machines that boost the performance of the in-house developed advanced seismic imaging algorithms.
NVIDIA’s Tesla GPU-accelerated computing platforms have been instrumental in supporting Eni’s exploration activity, improving our ability to turn around advanced seismic imaging tasks in a shorter time and with a higher accuracy. GPU-accelerated computing is a key competitive advantage for Eni, enabling us to accelerate the time to market of our discoveries and to shorten our overall upstream cycle.”
