Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “CRAY”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 469 records · Page 26

GPU Lifetimes on Titan Supercomputer: Survival Analysis and Reliability

The Cray XK7 Titan was the top supercomputer system in the world for a long time and remained critically important throughout its nearly seven year life. It was an interesting machine from a reliability viewpoint as most of its power came from 18,688 GPUs whose operation was forced to execute three rework cycles, two on the GPU mechanical assembly and one on the GPU circuitboards. We write about the last rework cycle and a reliability analysis of over 100,000 years of GPU lifetimes during Titan’s 6-year-long productive period. Using time between failures analysis and statistical survival analysis techniques, we find that GPU reliability is dependent on heat dissipation to an extent that strongly correlates with detailed nuances of the cooling architecture and job scheduling. We describe the history, data collection, cleaning, and analysis and give recommendations for future supercomputing systems. We make the data and our analysis codes publicly available.

Ostrouchov, George↗

The Anatomy of Mr. Scan: A Dissection of Performance of an Extreme Scale GPU-Based Clustering Algorithm, In: 2014 5th Workshop on Latest Advances in Scalable Algorithms for Large-Scale Systems

The emergence of leadership class systems with GPU-equipped nodes has the potential to vastly increase the performance of existing distributed applications. However, the inclusion of GPU computation into existing extreme scale distributed applications can reveal scalability issues that were absent in the CPU version. The issues exposed in scaling by a GPU can become limiting factors to overall application performance. We developed an extreme scale GPU-based application to perform data clustering on multi-billion point datasets. In this application, called Mr. Scan, we ran into several of these performance limiting issues. Through the use of complete end-to-end benchmarking of Mr. Scan (measuring time from reading and distribution to final output), we were able to identify three major sources of real world performance issues: data distribution, GPU load balancing, and system specific issues such as start-up time. These issues comprised a vast majority of the run time of Mr. Scan. Data distribution alone accounted for 68% of the total run time of Mr. Scan when processing 6.5 billion points on Cray Titan at 8192 nodes. With improvements in these areas, we have been able able to cut total run time of Mr. Scan from 17.5 minutes to 8.3 minutes when clustering 6.5 billion points.

Computer Science↗

Enabling parallel simulation of large-scale HPC network systems

Here, with the increasing complexity of today’s high-performance computing (HPC) architectures, simulation has become an indispensable tool for exploring the design space of HPC systems—in particular, networks. In order to make effective design decisions, simulations of these systems must possess the following properties: (1) have high accuracy and fidelity, (2) produce results in a timely manner, and (3) be able to analyze a broad range of network workloads. Most state-of-the-art HPC network simulation frameworks, however, are constrained in one or more of these areas. In this work, we present a simulation framework for modeling two important classes of networks used in today’s IBM and Cray supercomputers: torus and dragonfly networks. We use the Co-Design of Multi-layer Exascale Storage Architecture (CODES) simulation framework to simulate these network topologies at a flit-level detail using the Rensselaer Optimistic Simulation System (ROSS) for parallel discrete-event simulation. Our simulation framework meets all the requirements of a practical network simulation and can assist network designers in design space exploration. First, it uses validated and detailed flit-level network models to provide an accurate and high-fidelity network simulation. Second, instead of relying on serial time-stepped or traditional conservative discrete-event simulations that limit simulation scalability and efficiency, we use the optimistic event-scheduling capability of ROSS to achieve efficient and scalable HPC network simulations on today’s high-performance cluster systems. Third, our models give network designers a choice in simulating a broad range of network workloads, including HPC application workloads using detailed network traces, an ability that is rarely offered in parallel with high-fidelity network simulations

97 MATHEMATICS AND COMPUTING↗

Dynamic Adaptable Asynchronous Progress Model for MPI RMA Multiphase Applications

Casper is a process-based asynchronous progress model for MPI one-sided communication on multi-and many-core architectures. The one-sided communication is not truly one-sided in most MPI implementations: the target process still relies on software progress to complete incoming operations. Casper allows the user to specify an arbitrary number of cores dedicated to background ghost processes and transparently redirects the RMA operations to ghost processes by utilizing the PMPI redirection and MPI-3 shared-memory technologies. Although Casper benefits applications that suffer from lack of asynchronous progress, the operation redirection design might not support complex multiphase applications effectively, which often involve dynamically changing communication density and computing workloads. In this paper, we present an adaptive mechanism in Casper to address the limitation of static asynchronous progress in multiphase applications. For this we exploit two adaptive strategies, a user-guided strategy and a fully transparent and automatic strategy based on self-profiling and prediction, to dynamically reconfigure the asynchronous progress in Casper according to real-time performance characteristics during multiphase execution. We evaluate the adaptive approaches in both microbenchmarks and a real quantum chemistry application suite, NWChem, on the Cray XC30 supercomputer and an Intel Omni-Path cluster.

97 MATHEMATICS AND COMPUTING↗

Geometric Mapping of Tasks to Processors on Parallel Computers with Mesh or Torus Networks

We present a new method for reducing parallel applications' communication time by mapping their MPI tasks to processors in a way that lowers the distance messages travel and the amount of congestion in the network. Assuming geometric proximity among the tasks is a good approximation of their communication interdependence, we use a geometric partitioning algorithm to order both the tasks and the processors, assigning task parts to the corresponding processor parts. In this way, interdependent tasks are assigned to “nearby” cores in the network. We also present a number of algorithmic optimizations that exploit specific features of the network or application to further improve the quality of the mapping. We specifically address the case of sparse node allocation, where the nodes assigned to a job are not necessarily located in a contiguous block nor within close proximity to each other in the network. Furthermore, our methods generalize to contiguous allocations as well, and results are shown for both contiguous and non-contiguous allocations. We show that, for the structured finite difference mini-application MiniGhost, our mapping methods reduced communication time up to 75% relative to MiniGhost's default mapping on 128K cores of a Cray XK7 with sparse allocation. For the atmospheric modeling code E3SM/HOMME, our methods reduced communication time up to 31% on 16K cores of an IBM BlueGene/Q with contiguous allocation.

97 MATHEMATICS AND COMPUTING↗

Improving MPI Collective I/O for High Volume Non-Contiguous Requests With Intra-Node Aggregation

Two-phase I/O is a well-known strategy for implementing collective MPI-IO functions. It redistributes I/O requests among the calling processes into a form that minimizes the file access costs. As modern parallel computers continue to grow into the exascale era, the communication cost of such request redistribution can quickly overwhelm collective I/O performance. In this study, this effect has been observed from parallel jobs that run on multiple compute nodes with a high count of MPI processes on each node. Furthermore, to reduce the communication cost, we present a new design for collective I/O by adding an extra communication layer that performs request aggregation among processes within the same compute nodes. This approach can significantly reduce inter-node communication contention when redistributing the I/O requests. We evaluate the performance and compare it with the original two-phase I/O on Cray XC40 parallel computers (Theta and Cori) with Intel KNL and Haswell processors. Using I/O patterns from two large-scale production applications and an I/O benchmark, we show our proposed method effectively reduces the communication cost and hence maintains the scalability for a large number of processes.

97 MATHEMATICS AND COMPUTING↗

P3DFFT: A Framework for Parallel Computations of Fourier Transforms in Three Dimensions

Fourier and related transforms are a family of algorithms widely employed in diverse areas of computational science, notoriously difficult to scale on high-performance parallel computers with a large number of processing elements (cores). This paper introduces a popular software package called P3DFFT which implements fast Fourier transforms (FFTs) in three dimensions in a highly efficient and scalable way. It overcomes a well-known scalability bottleneck of three-dimensional (3D) FFT implementations by using two-dimensional domain decomposition. Designed for portable performance, P3DFFT achieves excellent timings for a number of systems and problem sizes. On a Cray XT5 system P3DFFT attains 45% efficiency in weak scaling from 128 to 65,536 computational cores. Library features include Fourier and Chebyshev transforms, Fortran and C interfaces, in- and out-of-place transforms, uneven data grids, and single and double precision. P3DFFT is available as open source at http://code.google.com/p/p3dfft/. This paper discusses P3DFFT implementation and performance in a way that helps guide the user in making optimal choices for parameters of their runs. Read More: https://epubs.siam.org/doi/10.1137/11082748X

Mathematics↗

Segmental Refinement: A Multigrid Technique for Data Locality

In this paper, we investigate a domain decomposed multigrid technique, termed segmental refinement, for solving general nonlinear elliptic boundary value problems. We extend the method first proposed in 1994 by analytically and experimentally investigating its complexity. We confirm that communication of traditional parallel multigrid is eliminated on fine grids, with modest amounts of extra work and storage, while maintaining the asymptotic exactness of full multigrid. We observe an accuracy dependence on the segmental refinement subdomain size, which was not considered in the original analysis. Finally, we present a communication complexity analysis that quantifies the communication costs ameliorated by segmental refinement and report performance results with up to 64K cores on a Cray XC30.

97 MATHEMATICS AND COMPUTING↗

Embedded ensemble propagation for improving performance, portability, and scalability of uncertainty quantification on emerging computational architectures

In this study, quantifying simulation uncertainties is a critical component of rigorous predictive simulation. A key component of this is forward propagation of uncertainties in simulation input data to output quantities of interest. Typical approaches involve repeated sampling of the simulation over the uncertain input data, and can require numerous samples when accurately propagating uncertainties from large numbers of sources. Often simulation processes from sample to sample are similar and much of the data generated from each sample evaluation could be reused. We explore a new method for implementing sampling methods that simultaneously propagates groups of samples together in an embedded fashion, which we call embedded ensemble propagation. We show how this approach takes advantage of properties of modern computer architectures to improve performance by enabling reuse between samples, reducing memory bandwidth requirements, improving memory access patterns, improving opportunities for fine-grained parallelization, and reducing communication costs. We describe a software technique for implementing embedded ensemble propagation based on the use of C++ templates and describe its integration with various scientific computing libraries within Trilinos. We demonstrate improved performance, portability and scalability for the approach applied to the simulation of partial differential equations on a variety of CPU, GPU, and accelerator architectures, including up to 131,072 cores on a Cray XK7 (Titan).

97 MATHEMATICS AND COMPUTING↗

Avoiding Communication in Primal and Dual Block Coordinate Descent Methods

Primal and dual block coordinate descent methods are iterative methods for solving regularized and unregularized optimization problems. Distributed-memory parallel implementations of these methods have become popular in analyzing large machine learning datasets. However, existing implementations communicate at every iteration, which, on modern data center and supercomputing architectures, often dominates the cost of floating-point computation. Recent results on communication-avoiding Krylov subspace methods suggest that large speedups are possible by re-organizing iterative algorithms to avoid communication. We show how applying similar algorithmic transformations can lead to primal and dual block coordinate descent methods that only communicate every $s$ iterations---where $s$ is a tuning parameter---instead of every iteration for the regularized least-squares problem. We show that the communication-avoiding variants reduce the number of synchronizations by a factor of $s$ on distributed-memory parallel machines without altering the convergence rate and attain strong scaling speedups of up to $6.1\times$ over the “standard algorithm" on a Cray XC30 supercomputer.

97 MATHEMATICS AND COMPUTING↗

Parrallel Implementation of Fast Randomized Algorithms for Low Rank Matrix Decomposition

We analyze the parallel performance of randomized interpolative decomposition by de- composing low rank complex-valued Gaussian random matrices larger than 100 GB. We chose a Cray XMT supercomputer as it provides an almost ideal PRAM model permitting quick investigation of parallel algorithms without obfuscation from hardware idiosyncrasies. We obtain that on non-square matrices performance scales almost linearly with runtime about 100 times faster on 128 processors. We also verify that numerically discovered error bounds still hold on matrices two orders of magnitude larger than those previously tested.

Lucas, Andrew J.↗

A scalable method for ab initio computation of free energies in nanoscale systems. In: SC '09 Proceedings of the Conference on High Performance Computing Networking, Storage and Analysis

Calculating the thermodynamics of nanoscale systems presents challenges in the simultaneous treatment of the electronic structure, which determines the interactions between atoms, and the statistical fluctuations that become ever more important at shorter length scales. Here we present a highly scalable method that combines ab initio electronic structure techniques, we use the Locally Self-Consitent Multiple Scattering (LSMS) technique, with the Wang-Landau (WL) algorithm to compute free energies and other thermodynamic properties of nanoscale systems. The combined WL-LSMS code is targeted to the study of nanomagnetic systems that have anywhere from about one hundred to a few thousand atoms. The code scales very well on the Cray XT5 system at ORNL, sustaining 1.03 Petaflop/s in double precision on 147,464 cores.

Eisenbach, M.↗

Atomistic nanoelectronic device engineering with sustained performances up to 1.44 PFlop/s

We present a multi-dimensional, atomistic, quantum transport simulation approach to investigate the performances of realistic nanoscale transistors for various geometries and material systems. The central computation consists in solving the Schrödinger equation with open boundary conditions several thousand times. To do that, a Wave Function approach is used since it can be relatively easily parallelized. To further improve the computational efficiency, three additional levels of parallelization are identified, the work load is optimally balanced between the CPUs, computational interleaving is applied where possible, and a mixed precision scheme is introduced. Using two different device types, a high electron mobility and a band-to-band tunneling transistor, sustained performances up to 1.28 PFlop/s in double precision (55% of the peak performance) and 1.44 PFlop/s in mixed precision are reached on 221,400 cores on the CRAY-XT5 Jaguar at Oak Ridge National Lab.

Luisier, Mathieu↗

Scalable implementations of accurate excited-state coupled cluster theories: application of high-level methods to porphyrin based systems

The development of reliable tools for excited-state simulations is emerging as an extremely powerful computational chemistry tool for understanding complex processes in the broad class of light harvesting systems and optoelectronic devices. Over the last years we have been developing equation of motion coupled cluster (EOMCC) methods capable of tackling these problems. In this paper we discuss the parallel performance of EOMCC codes which provide accurate description of the excited-state correlation effects. Two aspects are discuss in details: (1) a new algorithm for the iterative EOMCC methods based on the novel task scheduling algorithms, and (2) parallel algorithms for the non-iterative methods describing the effect of triply excited configurations. We demonstrate that the most computationally intensive non-iterative part can take advantage of 210,000 cores of the Cray XT5 system at OLCF. In particular, we demonstrate the importance of non-iterative many-body methods for achieving experimental level of accuracy for several porphyrin-based system.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

An analysis of computational workloads for the ORNL Jaguar system. In: ICS '12 Proceedings of the 26th ACM international conference on Supercomputing

This study presents an analysis of science application workloads for the Jaguar Cray XT5 system during its tenure as a 2.3 petaflop supercomputer at Oak Ridge National Laboratory. Jaguar was the first petascale system to be deployed for open science and has been one of the world's top three supercomputers for six releases of the TOP500 list. Its workload is investigated here as a representative of the growing worldwide install base of petascale systems and also as a foreshadowing of science workloads to be expected for future systems. The study considers characteristics of the Jaguar workload such as resource utilization, typical job characteristics, most heavily used applications, application scaling and application usage patterns. Implications of these findings are considered for current petascale workflows and for exascale systems to be deployed later this decade.

Joubert, Wayne↗

HykSort: a new variant of hypercube quicksort on distributed memory architectures. In: ICS '13 Proceedings of the 27th international ACM conference on International conference on supercomputing

In this paper, we present HykSort, an optimized comparison sort for distributed memory architectures that attains more than 2× improvement over bitonic sort and samplesort. The algorithm is based on the hypercube quicksort, but instead of a binary recursion, we perform a k-way recursion in which the pivots are selected accurately with an iterative parallel select algorithm. The single-node sort is performed using a vectorized and multithreaded merge sort. The advantages of HykSort are lower communication costs, better load balancing, and avoidance of O(p)-collective communication primitives. We also present a staged communication samplesort, which is more robust than the original samplesort for large core counts. We conduct an experimental study in which we compare hypercube sort, bitonic sort, the original samplesort, the staged samplesort, and HykSort. We report weak and strong scaling results and study the effect of the grain size. It turns out that no single algorithm performs best and a hybridization strategy is necessary. As a highlight of our study, on our largest experiment on 262,144 AMD cores of the CRAY XK7 "Titan" platform at the Oak Ridge National Laboratory we sorted 8 trillion 32-bit integer keys in 37 seconds achieving 0.9TB/s effective throughput.

Sundar, Hari↗

Mr. Scan: extreme scale density-based clustering using a tree-based network of GPGPU nodes, In: SC '13 Proceedings of the International Conference on High Performance Computing, Networking, Storage and Analysis

Density-based clustering algorithms are a widely-used class of data mining techniques that can find irregularly shaped clusters and cluster data without prior knowledge of the number of clusters it contains. DBSCAN is the most well-known density-based clustering algorithm. We introduce our version of DBSCAN, called Mr. Scan, which uses a hybrid parallel implementation that combines the MRNet tree-based distribution network with GPGPU-equipped nodes. Mr. Scan avoids the problems of existing implementations by effectively partitioning the point space and by optimizing DBSCAN's computation over dense data regions. We tested Mr. Scan on both a geolocated Twitter dataset and image data obtained from the Sloan Digital Sky Survey. At its largest scale, Mr. Scan clustered 6.5 billion points from the Twitter dataset on 8,192 GPU nodes on Cray Titan in 17.3 minutes. All other parallel DBSCAN implementations have only demonstrated the ability to cluster up to 100 million points.

Computer Science↗

Taming parallel I/O complexity with auto-tuning

We present an auto-tuning system for optimizing I/O performance of HDF5 applications and demonstrate its value across platforms, applications, and at scale. The system uses a genetic algorithm to search a large space of tunable parameters and to identify effective settings at all layers of the parallel I/O stack. The parameter settings are applied transparently by the auto-tuning system via dynamically intercepted HDF5 calls. To validate our auto-tuning system, we applied it to three I/O benchmarks (VPIC, VORPAL, and GCRM) that replicate the I/O activity of their respective applications. We tested the system with different weak-scaling configurations (128, 2048, and 4096 CPU cores) that generate 30 GB to 1 TB of data, and executed these configurations on diverse HPC platforms (Cray XE6, IBM BG/P, and Dell Cluster). In all cases, the auto-tuning framework identified tunable parameters that substantially improved write performance over default system settings. In conclusion, we consistently demonstrate I/O write speedups between 2x and 100x for test configurations.

97 MATHEMATICS AND COMPUTING↗