No sooner had one system, used by BP, been declared the most powerful supercomputer in the industrial sphere than it quickly has been displaced. Last Thursday (Jan. 18) it was Italian energy company Eni’s turn to take the lead with the launch of its latest petawhopper. At 18.6 petaflops (peak) the new cluster, HPC4, becomes the world’s most powerful commercial system (that we know of) and quadruples the company’s computing capacity to an aggregate peak performance of 22.4 petaflops.
With the debut of HPC4, the baton for industrial HPC leadership is passed from BP to Eni. In December, BP boosted its top supercomputer’s processing speed to 9 petaflops. Total was the previous record-holder with the 6.7-petaflop “Pangea” system, introduced in 2016.
“HPC4 is an important achievement in Eni’s digitalisation process,” commented Eni CEO Claudio Descalzi. The new machine will host the ecosystem of algorithms developed by Eni to support its activities in the exploration and production sector. In the risky and competitive energy exploration space, operating the latest HPC tech is essential for high-accuracy, high-resolution seismic imaging, geological modelling and reservoir simulation.
“These technologies will enable us, on the one hand, to accelerate and make the entire upstream process more efficient and accurate, reducing risks in the exploration phase and, at the same time, giving us a significant technological advantage, but also to increase the level of reliability, technical integrity and operability of all our productive plants, while minimising operational risks, with benefits both in terms of safety and environmental impact,” said Descalzi.
If HPC4 were given a run at the current Top500 list, there’s a good chance it would crack the top ten, breaking new ground for a commercially-owned supercomputer. By the time the new system is benchmarked for the June list, a top 20 position is a safe bet.
HPC4 joins the Lenovo-built HPC3 cluster, which ranks 51st on the current Top500 list with 2.6 petaflops Linpack (3.8 petaflops peak). Designed by Hewlett Packard Enterprise (HPE), the new cluster encompasses 1,600 ProLiant DL380 nodes, linked with EDR InfiniBand. Each node is equipped with two Intel 24-core Skylake processors and two Nvidia Tesla P100 GPU accelerators. It also includes a 15 petabytes storage subsystem. The system has a purported pricetag of $25 million, according to a Bloomberg report.
Both systems will be housed inside Eni’s Green Data Center, located in Ferrera Erbognone in Pavia, Italy. Built on a former rice paddy, the Green Data Center opened in 2013 to host all of Eni’s HPC architecture and its business applications.
From the earliest days of supercomputing, the oil and gas industry has led the commercial sector in its demand for HPC. Energy companies were the first to cross the petascale threshold for industry and now they are looking to the exascale horizon.
With competitive pressures mounting to process ever-increasing amounts of data to accelerate cycle times in the upstream process, Eni has its eye on exascale. “With HPC4 we are tracing the path for the use of exascale supercomputers in the energy sector that could revolutionise the way in which oil & gas activities are managed,” said CEO Descalzi.
Based in Italy, Eni is active in 73 countries and has more than 32,000 employees. The company’s exploration and production (E&P) division, which is focused on finding and producing oil and gas, has a presence in 42 countries.