New Frontier in High-Performance Computing: Reservoir Model with Billion Cells

Abstract The oil and gas industry increasingly relies on high-resolution simulation to improve understanding of complex geological systems and inform decision-making at field and regional scales. This paper presents a major milestone in reservoir modeling: the construction and execution of a billion-cell dynamic reservoir model, marking a new frontier in high-performance computing (HPC) applications for subsurface simulation. Earlier modeling efforts (Bogachev 2024) were constrained not by memory architecture itself, but by internal indexing limits associated with 32-bit integers - specifically, the maximum number of grid blocks addressable using the standard int type. In this work, we transitioned to 64-bit indexing throughout the simulation workflow, allowing support for grids of 10 billion cells. This upgrade, along with optimized data storage formats and enhanced parallel load balancing, enabled the successful deployment of a fully integrated static and dynamic model on a compute cluster. These improvements resolve long-standing bottlenecks related to grid scalability, file Input/Output (I/O) size, and execution efficiency. The model represents a research-grade, full-field compositional simulation derived from a real geological setting, designed to emulate a regional system. By preserving fine-scale resolution across vast spatial domains, it captures inter-reservoir connectivity and large-scale flow dynamics often lost in conventional, upscaled models. This capability is critical for regional planning, exploration strategy, and long-term resource forecasting. This study underscores the convergence of advanced software design, cluster-native workflows, and scalable computation to unlock next-generation reservoir simulation. The findings offer practical pathways for integrating geological complexity without sacrificing computational feasibility, paving the way for full-scale modeling of megastructures and multi-reservoir systems in real-world asset development.