Abstract
Across many fields of science and engineering computers now play a significant role in scientific discovery through both large scale simulation and real time data acquisition. As these scientific simulations increasingly require new levels of complexity and fidelity, and leveraging increasing computational capabilities, scientists are migrating their applications back and forth from workstations to supercomputers, high performance clusters, and new distributed grids of computers. For applications which depend on live real world data, the migration to varied and distributed resources provides additional challenges. This migration involves reassigning and testing individual sensor communications and data path integrity for the application before the application is ready for a real time operating environment.
This paper presents one general approach for communicating live real world data to simulations deployed on high-end computational resources. The architecture and design are presented for generic applications, before focusing on a particular scenario for drilling dynamics optimizations. In this scenario, simulations of drilling behavior depend on real time data from on-site (remote) sensors. Sensor data is streamed through to a simulation framework. The framework (available on a desktop computer or a supercomputer) displays relevant information, starts simulations with real time inputs then presents results and recommendations. For our application, we will use the Cactus Code (www.cactuscode.org) open-source high performance scientific computing framework as the simulation framework, and LabVIEW (www.ni.com/LabVIEW) as the data acquisition software. The advances in architecture portability allowed by the framework are summarized, and experiences of system uses are presented. Discussion will include opportunities found and learnings from using this platform in various environments highlighting drilling optimization.
