A Toolbox of Scalable Algorithms and Software for On-Line and Off-Line Management of Advanced Scientific Simulations

A central problem for many large-scale scientific and engineering simulations is the efficient archival and analysis of the tremendous amount of data that is generated. To address this problem, in this paper we present algorithms for the efficient compression and subsequent querying of such data sets on parallel computers. In addition, we discuss the integration of these algorithms into a software system for the on-line monitoring, steering, and visualization of such simulations. This system, SAVE is based on a functional representation of the underlying simulation data and the use of POSSE—a portable, object-oriented, scientific steering environment. We present experimental results for the use of this system within a computational fluid dynamics simulation of a turbulent reacting flow. We believe that this software represents a promising tool for the on-line and off-line management of large-scale scientific simulation data.