By embracing dense GPU systems, modern algorithms, and software design approaches, ECHELON software achieves performance levels that significantly outpace the competition.
For the month of April, I am fulfilling a long-time dream of living in Florence and learning the Italian language. I’ve been thinking about the vastly different world it was when my grandparents left Italy in the 1930s. I arrived comfortably by air in about seven hours, while it took weeks for them to cross the ocean on a ship. Vast technological advances allow me to communicate real-time with my family and my office essentially at will from almost anywhere in the world. However, a mere two generations back, it took months for letters to cross the Atlantic. There is great value in speed and in embracing the technology that enables it, as the benefits can be both disruptive and transformative to our life experience.
In the field of reservoir simulation, I think of Stone Ridge Technology’s ECHELON software as one such disruptive technology. Over the last two decades there have been numerous advances in hardware, software and algorithms. In hardware, for example, processor clock speeds plateaued more than a decade ago and chip vendors shifted to multi- and now many-core architectures. Consequently, advances in software performance are now only achieved through a complex hierarchy of parallelism from coarse-grained domain decomposition down to fine-grained, e.g., thread-level, instruction level and data parallelism.
In addition, GPUs have joined CPUs as general-purpose compute platforms and have advanced briskly, outpacing CPU hardware in both memory bandwidth and FLOPS, both of which are critical performance factors for reservoir simulation. ECHELON is the first reservoir simulation software written explicitly for GPUs to fully employ both coarse-grain and all levels of fine-grain parallelism available. Every computational kernel runs on GPUs. ECHELON also uses optimal solver algorithms instead of employing those that are more easily parallelized but algorithmically inefficient. The software was developed from scratch using a modern high-performance object-oriented software design approach to allow the quick addition of new and evolving engineering features.
By embracing newer, dense GPU systems, modern algorithms and software design approaches, ECHELON software achieves performance levels that significantly outpace the competition. When compared to CPU-based hardware and software alternatives, ECHELON software is exceptionally fast, — ranging from 10 to more than 100 times faster on real field assets containing complex engineering features. Furthermore, looking at the GPU technology roadmap recently announced in early April at NVIDIA GTC 2016, the performance gap between ECHELON and its CPU-bound competitors will widen in the coming years.
We have seen that the combination of ECHELON software and the Cray® CS-Storm system’s dense GPU platform makes a compelling solution for organizations focused on increasing the productivity of their field assets, in an economic climate that demands lower costs. The value of speed, the ability to model at the geo-scale and the ability to employ more compact, powerful hardware have emerged as salient requirements in the E&P industry. Running on Cray/NVIDIA hardware, ECHELON software can enhance productivity and lower costs for companies in several ways:
With this solution, Stone Ridge Technology, Cray and NVIDIA can help exploration and production companies boost productivity and lower costs at a critical time in the industry.
Vincent Natoli is the president and founder of Stone Ridge Technology. He is a computational physicist with 30 years experience in the field of high-performance computing. He holds Bachelors and Masters degrees from MIT, a PhD in Physics from the University of Illinois Urbana-Champaign and a Masters in Technology Management from the Wharton School at the University of Pennsylvania.
What we are doing to help improve the reservoir simulation industry.
Leveraging modern GPUs and neural computing libraries, scientists are using deep learning technology to solve differential equations that emerge from a diverse set of physics-based problems. This blog introduces the topic and explores potential applications and limitations of the technology.Read article →