arXiv subjects

Scott Klasky

Publications and source records attributed to Scott Klasky.

44 records · Page 3Linked to original sources

arXiv2021

Scalable Multigrid-based Hierarchical Scientific Data Refactoring on GPUs

Rapid growth in scientific data and a widening gap between computational speed and I/O bandwidth makes it increasingly infeasible to store and share all data produced by scientific simulations. Instead, we need methods for reducing data volumes: ideally, methods that can scale data volumes adaptively so as to enable negotiation of performance and fidelity tradeoffs in different situations. Multigrid-based hierarchical data representations hold promise as a solution to this problem, allowing for flexible conversion between different fidelities so that, for example, data can be created at high fidelity and then transferred or stored at lower fidelity via logically simple and mathematically sound operations. However, the effective use of such representations has been hindered until now by the relatively high costs of creating, accessing, reducing, and otherwise operating on such representations. We describe here highly optimized data refactoring kernels for GPU accelerators that enable efficient creation and manipulation of data in multigrid-based hierarchical forms. We demonstrate that our optimized design can achieve up to 264 TB/s aggregated data refactoring throughput -- 92% of theoretical peak -- on 1024 nodes of the Summit supercomputer. We showcase our optimized design by applying it to a large-scale scientific visualization workflow and the MGARD lossy compression software.

cs.DC↗

arXiv2021

Accelerating Multigrid-based Hierarchical Scientific Data Refactoring on GPUs

Rapid growth in scientific data and a widening gap between computational speed and I/O bandwidth make it increasingly infeasible to store and share all data produced by scientific simulations. Instead, we need methods for reducing data volumes: ideally, methods that can scale data volumes adaptively so as to enable negotiation of performance and fidelity tradeoffs in different situations. Multigrid-based hierarchical data representations hold promise as a solution to this problem, allowing for flexible conversion between different fidelities so that, for example, data can be created at high fidelity and then transferred or stored at lower fidelity via logically simple and mathematically sound operations. However, the effective use of such representations has been hindered until now by the relatively high costs of creating, accessing, reducing, and otherwise operating on such representations. We describe here highly optimized data refactoring kernels for GPU accelerators that enable efficient creation and manipulation of data in multigrid-based hierarchical forms. We demonstrate that our optimized design can achieve up to 250 TB/s aggregated data refactoring throughput -- 83% of theoretical peak -- on 1024 nodes of the Summit supercomputer. We showcase our optimized design by applying it to a large-scale scientific visualization workflow and the MGARD lossy compression software.

cs.DC↗

arXiv2020

MGARD+: Optimizing Multilevel Methods for Error-bounded Scientific Data Reduction

Data management is becoming increasingly important in dealing with the large amounts of data produced by large-scale scientific simulations and instruments. Existing multilevel compression algorithms offer a promising way to manage scientific data at scale, but may suffer from relatively low performance and reduction quality. In this paper, we propose MGARD+, a multilevel data reduction and refactoring framework drawing on previous multilevel methods, to achieve high-performance data decomposition and high-quality error-bounded lossy compression. Our contributions are four-fold: 1) We propose a level-wise coefficient quantization method, which uses different error tolerances to quantize the multilevel coefficients. 2) We propose an adaptive decomposition method which treats the multilevel decomposition as a preconditioner and terminates the decomposition process at an appropriate level. 3) We leverage a set of algorithmic optimization strategies to significantly improve the performance of multilevel decomposition/recomposition. 4) We evaluate our proposed method using four real-world scientific datasets and compare with several state-of-the-art lossy compressors. Experiments demonstrate that our optimizations improve the decomposition/recomposition performance of the existing multilevel method by up to 70X, and the proposed compression method can improve compression ratio by up to 2X compared with other state-of-the-art error-bounded lossy compressors under the same level of data distortion.

cs.DC↗

arXiv2020

Towards 1ULP evaluation of Daubechies Wavelets

We present algorithms to numerically evaluate Daubechies wavelets and scaling functions to high relative accuracy. These algorithms refine the suggestion of Daubechies and Lagarias to evaluate functions defined by two-scale difference equations using splines; carefully choosing amongst a family of rapidly convergent interpolators which effectively capture all the smoothness present in the function and whose error term admits a small asymptotic constant. We are also able to efficiently compute derivatives, though with a smoothness-induced reduction in accuracy. An implementation is provided in the Boost Software Library.

math.NA↗

arXiv2018

A tight-coupling scheme sharing minimum information across a spatial interface between gyrokinetic turbulence codes

A new scheme that tightly couples kinetic turbulence codes across a spatial interface is introduced. This scheme evolves from considerations of competing strategies and down-selection. It is found that the use of a composite kinetic distribution function and fields with global boundary conditions as if the coupled code were one, makes the coupling problem tractable. In contrast, coupling the two solutions from each code across the overlap region is found to be more difficult due to numerical dephasing of the turbulent solutions between two solvers. Another advantage of the new scheme is that the data movement can be limited to the 3D fluid quantities, instead of higher dimensional kinetic information, which is computationally more efficient for large scale simulations on leadership class computers.

physics.plasm-ph↗

arXiv2017

On the Scalability of Data Reduction Techniques in Current and Upcoming HPC Systems from an Application Perspective

We implement and benchmark parallel I/O methods for the fully-manycore driven particle-in-cell code PIConGPU. Identifying throughput and overall I/O size as a major challenge for applications on today's and future HPC systems, we present a scaling law characterizing performance bottlenecks in state-of-the-art approaches for data reduction. Consequently, we propose, implement and verify multi-threaded data-transformations for the I/O library ADIOS as a feasible way to trade underutilized host-side compute potential on heterogeneous systems for reduced I/O latency.

cs.PF↗

arXiv2016

Towards Real-Time Detection and Tracking of Spatio-Temporal Features: Blob-Filaments in Fusion Plasma

A novel algorithm and implementation of real-time identification and tracking of blob-filaments in fusion reactor data is presented. Similar spatio-temporal features are important in many other applications, for example, ignition kernels in combustion and tumor cells in a medical image. This work presents an approach for extracting these features by dividing the overall task into three steps: local identification of feature cells, grouping feature cells into extended feature, and tracking movement of feature through overlapping in space. Through our extensive work in parallelization, we demonstrate that this approach can effectively make use of a large number of compute nodes to detect and track blob-filaments in real time in fusion plasma. On a set of 30GB fusion simulation data, we observed linear speedup on 1024 processes and completed blob detection in less than three milliseconds using Edison, a Cray XC30 system at NERSC.

cs.DC↗

arXiv2014

Region Templates: Data Representation and Management for Large-Scale Image Analysis

Distributed memory machines equipped with CPUs and GPUs (hybrid computing nodes) are hard to program because of the multiple layers of memory and heterogeneous computing configurations. In this paper, we introduce a region template abstraction for the efficient management of common data types used in analysis of large datasets of high resolution images on clusters of hybrid computing nodes. The region template provides a generic container template for common data structures, such as points, arrays, regions, and object sets, within a spatial and temporal bounding box. The region template abstraction enables different data management strategies and data I/O implementations, while providing a homogeneous, unified interface to the application for data storage and retrieval. The execution of region templates applications is coordinated by a runtime system that supports efficient execution in hybrid machines. Region templates applications are represented as hierarchical dataflow in which each computing stage may be represented as another dataflow of finer-grain tasks. A number of optimizations for hybrid machines are available in our runtime system, including performance-aware scheduling for maximizing utilization of computing devices and techniques to reduce impact of data transfers between CPUs and GPUs. An experimental evaluation on a state-of-the-art hybrid cluster using a microscopy imaging study shows that this abstraction adds negligible overhead (about 3%) and achieves good scalability.

cs.DC↗