Advanced modeling and simulation has enabled the design of a variety of innovative products and the analysis of numerous complex phenomenon. However, significant barriers exist to widespread adoption of these tools. In particular, advanced modeling and simulation: (1) is considered complex to use, (2) needs in-house expertise, and (3) requires high capital costs. In this paper, we describe the development of an end-to-end, advanced modeling and simulation cloud platform that encapsulates best practices for scientific computing in the cloud, and demonstrate using Hydra-TH as a prototypical application. As an alternative to traditional advanced modeling and simulation workflows, our Web-based approach simplifies the processes, decreases the need for in-house computational science and engineering experts, and lowers the capital investments. In addition to providing significantly improved, intuitive software, the environment offers reproducible workflows where the full lifecycle of data from input to final analyzed results can be saved, shared, and even published.

PDS/PIO is a lightweight, parallel interface designed to support efficient transfers of massive, grid-based, simulation data among memory, disk, and tape subsystems. The higher-level PDS (Parallel Data Set) interface manages data with tensor and unstructured grid abstractions, while the lower-level PIO (Parallel Input/Output) interface accesses data arrays with arbitrary permutation, and provides communication and collective I/O operations. Higher-level data abstraction for finite element applications is provided by PXI (Parallel Exodus Interface), which supports, in parallel, functionality of Exodus II, a finite element data model developed at Sandia National Laboratories. The entire interface is implemented in C with Fortran-callable PDS and PXI wrappers.

Building complex meshes for large-scale numerical simulations presents immense difficulties in exploiting high-performance computers. Industry and research leaders describe the current state of the art for generating meshes for such large scientific problems. This is followed by a panel and general audience discussion of the algorithmic and architectural issues surrounding the generation of meshes with 10/sup 7/ to 10/sup 9/ grid points. (The terms "mesh" and "grid" are used interchangeably in the literature).>

The implementation of a fully vectorized 3-D finite-element model on the Cray X-MP Y-MP architecture is presented. Issues involved in vectorizing the computationally intensive modules of the finite-element model are discussed. Fully vectorized numerical integration coupled with a preconditioned conjugate gradient linear system solver results in an application performance of 140 MFLOPS, with the linear system solver achieving 162 MFLOPS (Y-MP single processor). The vector/scalar speedup for the finite-element code is approximately 14.>

Fully vectorized versions of the Los Alamos National Laboratory benchmark code Gamteb, a Monte Carlo photon transport algorithm, were developed for the Cyber 205/ETA-10 and Cray X-MP/Y-MP architectures. Single-processor performance measurements of the vector and scalar implementations were modeled in a modified Amdahl's Law that accounts for additional data motion in the vector code. The performance and implementation strategy of the vector codes are related to architectural features of each machine. Speedups between fifteen and eighteen for Cyber 205/ETA-10 architectures, and about nine for CRAY X-MP/Y-MP architectures are observed. The best single processor execution time for the problem was 0.33 seconds on the ETA-10G, and 0.42 seconds on the CRAY Y-MP.