The oil and gas industry can more effectively perform reservoir simulation, leading to more accurate evaluation of both new and old reservoirs thanks to bench-marking completed by Cray using its CPU/GPU based CS-Storm supercomputers.
While the oil and gas industry faces declining oil prices worldwide, it must look to increase the efficiency of its operations to maintain profits. One way to do this is with high quality simulation tools, which can provide more precise information on the potential oil and gas reserves in a given reservoir.
The project from Cray focused on the Echelon reservoir simulation software from Stone Ridge Technology, which provides services and products to the HPC industry specifically targeting Oil and Gas. Echelon is used to model oil, water and gas flows in a reservoir, and is specifically designed for GPU accelerated compute architectures. Creating code specifically for GPUs rather than retroactively adapting it can provide a significant advantage as the application can take advantage of the potential for parallelism in the GPU hardware more effectively, increasing simulation speed, reducing the time taken to process simulations and derive meaningful results.
Bert Beals, director for Cray’s Energy segment said: ‘With Echelon and the Cray CS-Storm system, reservoir engineers now have the ability to run multiple realisations in a single day. Performance requirements in the oil and gas industry, and particularly around reservoir simulation, are driven by demand for new oil fields and evaluation of existing fields for future investment. Speed, reliability and scalability are all important factors, and Echelon running on our GPU-packed Cray CS-Storm system will provide oil and gas customers with a powerful solution for successfully evaluating the potential of both new and old reservoirs.
Echelon software was specifically designed around the use of GPUs. By running this application on the Cray CS-Storm system, Echelon can run high-resolution reservoir models more rapidly, the combined technologies of Cray and Stone Ridge can deliver higher degrees of accuracy, without compromising production schedules.
‘Runtime performance is absolutely vital in the oil and gas industry, playing a critical role in reservoir simulation,’ said Vincent Natoli, Founder and CEO of Stone Ridge Technology. ‘We have demonstrated results on a 243 million-cell model running out 45 years on just two CS-Storm nodes completing in about two-and-a-half hours. The ability to do more in a smaller hardware footprint brings significant and welcome cost savings and productivity improvements at a very competitive time in the energy industry’ concluded Natoli.
The Cray CS-Storm is targeting applications that can take advantage of highly parallel GPU heavy compute architecture such as in areas such as energy, manufacturing, earth sciences, life sciences and financial services. Each CS-Storm chassis combines eight NVIDIA Tesla K40 or K80 GPU accelerators with Intel Xeon processors. A single chassis equipped with Tesla K80s delivers 15 teraflops of peak performance, while a full petaflop requires just three racks.
The CS-Storm system is designed around GPU based computing, providing reliable operation while driving all eight 300W GPUs at maximum performance. The system is available with a unique Cray Programming Environment, enabling users to leverage the parallelism available from GPUs, and sophisticated systems management for lights-out computing.
