Rupak Biswas

Memory-intensive benchmarks: IRAM vs. cache-based machines

The increasing gap between processor and memory performance has led to new architectural models for memory-intensive applications. In this paper, we use a set of memory-intensive benchmarks to evaluate a mixed logic and DRAM processor called VIRAM as a building block for scientific computing. For each benchmark, we explore the fundamental hardware requirements of the problem as well as alternative algorithms and data structures that can help expose fine-grained parallelism or simplify memory access patterns. Results indicate that VIRAM is significantly faster than conventional cache-based machines for problems that are truly limited by the memory system and that it has a significant power advantage across all the benchmarks.

An Overset Grid Navier-Stokes/Kirchhoff-Surface Method for Rotorcraft Aeroacoustic Predictions

Abstract : This paper describes a new method for computing the flowfield and acoustic signature of arbitrary rotors in forward flight. The overall scheme uses a finite-difference Navier-Stokes solver to compute the aerodynamic flowfield near the rotor blades. The equations are solved on a system of overset grids that allow for prescribed cyclic and flapping blade motions and capture the interactions between the rotor blades and wake. The far-field noise is computed with a Kirchhoff integration over a surface that completely encloses the rotor blades. FIowfield data are interpolated onto this Kirchhoff surface using the same overset-grid techniques that are used for the flowfield solution. As a demonstration of the overall prediction scheme, computed results for far-field noise are compared with experimental data for both high-speed impulsive (HSI) and blade-vortex interaction (BVI) cases. The HS! case showed good agreement with experimental data while a preliminary attempt at the BVI case did not. The computations clearly show that temporal accuracy, spatial accuracy and grid resolution in the Navier-Stokes solver play key roles in the overall accuracy of the predicted noise. These findings will be addressed more closely in future BVI computations.

A dynamic mesh adaption procedure for unstructured hexahedral grids

Hexahedral elements can be subdivided anisotropically without mesh quality problems that are associated with tetrahedral meshes. Furthermore, hexahedral meshes yield more accurate solutions than their tetrahedral counterparts for the same number of edges. Our adaption procedure uses an edge data structure that facilitates efficient subdivision by allowing individual edges to be marked for refinement or coarsening. Pyramids and prisms are used as buffer elements between refined and unrefined hexahedra to eliminate hanging vertices. Preliminary results indicate that this new adaption procedure is a viable alternative to adaptive tetrahedral schemes. (Author)

5. Performance Evaluation and Modeling of Ultra-Scale Systems

The growing gap between sustained and peak performance for full-scale complex scientific applications on conventional supercomputers is a major concern in high performance computing (HPC). The problem is expected to be exacerbated by the end of this decade, as mission-critical applications will have computational requirements that are at least two orders of magnitude larger than current levels. In order to continuously increase raw computational power and at the same time substantially reap its benefits, major strides are necessary in hardware architecture, software infrastructure, and application development. The first step toward this goal is the accurate assessment of existing and emerging HPC systems across a comprehensive set of scientific algorithms. In addition, high-fidelity performance modeling is required to understand and predict the complex interactions among hardware, software, and applications, and thereby influence future design trade-offs. This survey article discusses recent performance evaluations of state-of-the-art ultra-scale systems for a diverse set of scientific applications, including scalable compact synthetic benchmarks and architectural probes. In addition, performance models and program characterizations from key scientific areas are described.

Scheduling in Heterogeneous Grid Environments: The Effects of DataMigration

Computational grids have the potential for solving large-scale scientific problems using heterogeneous and geographically distributed resources. However, a number of major technical hurdles must be overcome before this goal can be fully realized. One problem critical to the effective utilization of computational grids is efficient job scheduling. Our prior work addressed this challenge by defining a grid scheduling architecture and several job migration strategies. The focus of this study is to explore the impact of data migration under a variety of demanding grid conditions. We evaluate our grid scheduling algorithms by simulating compute servers, various groupings of servers into sites, and inter-server networks, using real workloads obtained from leading supercomputing centers. Several key performance metrics are used to compare the behavior of our algorithms against reference local and centralized scheduling schemes. Results show the tremendous benefits of grid scheduling, even in the presence of input/output data migration - while highlighting the importance of utilizing communication-aware scheduling schemes.Computational grids have the potential for solving large-scale scientific problems using heterogeneous and geographically distributed resources. However, a number of major technical hurdles must be overcome before this potential can be realized. One problem that is critical to effective utilization of computational grids is the efficient scheduling of jobs. This work addresses this problem by describing and evaluating a grid scheduling architecture and three job migration algorithms. The architecture is scalable and does not assume control of local site resources. The job migration policies use the availability and performance of computer systems, the network bandwidth available between systems, and the volume of input and output data associated with each job. An extensive performance comparison is presented using real workloads from leading computational centers. The results, based on several key metrics, demonstrate that the performance of our distributed migration algorithms is significantly greater than that of a local scheduling framework and comparable to a non-scalable global scheduling approach.