How Innovations In High Performance Computing and Visualization Can Benefit Reservoir Characterization
Zollie L. Johnson, Melinda K. Manning · Offshore Technology Conference · 1997
Abstract Today, most E & P research makes use of large, three dimensional data volumes to model, simulate and solve real world problems. This is especially evident when working with reservoir characterization and simulation data. Increasingly, there are problems that require up to many terabytes of data per run. Data sizes are increasing by orders of magnitude as teraflop scalable supercomputers become available as standard E & P machines. As a scalable supercomputing resource, these systems can be configured to meet computing needs ranging from the smallest to the largest, most complex and demanding reservoir problem. In the past few years, the cost of supercomputing technology has decreased while the performance has increased. This has allowed E & P companies to deploy this type of technology throughout their organization in places never before possible. For example, it is now possible to accommodate remote sites with smaller supercomputer systems and central engineering sites with larger systems. This enables cost control, while providing supercomputing resources at the locations where they are needed the most. Scalability includes adding computing resources as required: more parallel inputloutput (UO) with faster I/O performance, more parallel processing units for compute intensive problems. large distributed shared memories for terabyte size data sets and high-performance graphics visualization. Highly scalable supercomputers can beused as a tool to better predict the behavior of the reservoir and produce more oil and gas from known reserves. With unprecedented graphics and computing power, these machines compute and visualize the complex relationships among parameters that must be considered before. during, and after a simulation run. This visual data integration enables users to better understandthe relationships among fluid and rock properties within the reservoir, both before and after simulation. Users are able to optimize simulation time by making decisions based on a more complete and thorough understanding of data through visualization. This paper will show how innovations in high performance computing and visualization can benefit work inreservoir characterization by discussing these significant shifts in computing technology. 1.0 Innovations In Computer Architecture A common requirement throughout the industry is to be able to collect and process large data volumes that will result into a realistic representation of the sub-surface. It is becoming more common for geoscientists to visualize datasets that are a gigabyte in size while some companies are now requesting for methods to process and selectively visualize terabyte size data sets, Innovations which allow this type of processing and visualization to occur are inherent in the design of the Scalable Shared-Memory Multiprocessor (S2MP) system architecture. Figure 1. Shows the S2MP architecture which utilizes the following features:Distributed Shared-Memory and I/OSystem InterconnectionsDirectory-based Cache CoherencePage Migration and Replication Distributed shared-memory (DSM) and I/O : S2MP memory is physically dispersed throughout the system for faster processor access. Although main memory is distributed, it is universally accessible and shared between all the processors in the system. Similarly. I/O devices are distributed among the nodes, and each device is accessible to every processor in the system.