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 343 records · Page 19

Integrating D3 Technology into an Existing DMF Configuration

The Numerical Aerodynamics Simulation (NAS) Facility's CRAY C916/1024 has 240+ GB of SuperHome storage with 2 TB of migrated storage. NAS is a beta test site for the Storage Technology Corporation (STK) Redwood (SD-3) device and Cray's DMF 2.3. This discussion will include a list of resolved problems and various metrics relative to transfer, search, rewind, unload, and mount timings. Also covered will be how DMF 2.3 with the advance tape MSP will integrate the Redwood tape drives along with the existing STK 4490 tape drives to potentially reduce the number of tapes used by 85%.

Powers, Alan K.↗

Achieving High Performance on the i860 Microprocessor

The i860 is a high performance microprocessor used in the Intel Touchstone project. This paper proposes a paradigm for programming the i860 that is modelled on the vector instructions of the Cray computers. Fortran callable assembler subroutines were written that mimic the concurrent vector instructions of the Cray. Cache takes the place of vector registers. Using this paradigm we have achieved twice the performance of compiled code on a traditional solve.

Lee, King↗

Performance Analysis of the NAS Y-MP Workload

This paper describes the performance characteristics of the computational workloads on the NAS Cray Y-MP machines, a Y-MP 832 and later a Y-MP 8128. Hardware measurements indicated that the Y-MP workload performance matured over time, ultimately sustaining an average throughput of 0.8 GFLOPS and a vector operation fraction of 87%. The measurements also revealed an operation rate exceeding 1 per clock period, a well-balanced architecture featuring a strong utilization of vector functional units, and an efficient memory organization. Introduction of the larger memory 8128 increased throughput by allowing a more efficient utilization of CPUs. Throughput also depended on the metering of the batch queues; low-idle Saturday workloads required a buffer of small jobs to prevent memory starvation of the CPU. UNICOS required about 7% of total CPU time to service the 832 workloads; this overhead decreased to 5% for the 8128 workloads. While most of the system time went to service I/O requests, efficient scheduling prevented excessive idle due to I/O wait. System measurements disclosed no obvious bottlenecks in the response of the machine and UNICOS to the workloads. In most cases, Cray-provided software tools were- quite sufficient for measuring the performance of both the machine and operating, system.

Bergeron, Robert J.↗

Parallelization of ARC3D with Computer-Aided Tools

A series of efforts have been devoted to investigating methods of porting and parallelizing applications quickly and efficiently for new architectures, such as the SCSI Origin 2000 and Cray T3E. This report presents the parallelization of a CFD application, ARC3D, using the computer-aided tools, Cesspools. Steps of parallelizing this code and requirements of achieving better performance are discussed. The generated parallel version has achieved reasonably well performance, for example, having a speedup of 30 for 36 Cray T3E processors. However, this performance could not be obtained without modification of the original serial code. It is suggested that in many cases improving serial code and performing necessary code transformations are important parts for the automated parallelization process although user intervention in many of these parts are still necessary. Nevertheless, development and improvement of useful software tools, such as Cesspools, can help trim down many tedious parallelization details and improve the processing efficiency.

Jin, Haoqiang↗

Parallel Calculation of Sensitivity Derivatives for Aircraft Design using Automatic Differentiation

Sensitivity derivative (SD) calculation via automatic differentiation (AD) typical of that required for the aerodynamic design of a transport-type aircraft is considered. Two ways of computing SD via code generated by the ADIFOR automatic differentiation tool are compared for efficiency and applicability to problems involving large numbers of design variables. A vector implementation on a Cray Y-MP computer is compared with a coarse-grained parallel implementation on an IBM SP1 computer, employing a Fortran M wrapper. The SD are computed for a swept transport wing in turbulent, transonic flow; the number of geometric design variables varies from 1 to 60 with coupling between a wing grid generation program and a state-of-the-art, 3-D computational fluid dynamics program, both augmented for derivative computation via AD. For a small number of design variables, the Cray Y-MP implementation is much faster. As the number of design variables grows, however, the IBM SP1 becomes an attractive alternative in terms of compute speed, job turnaround time, and total memory available for solutions with large numbers of design variables. The coarse-grained parallel implementation also can be moved easily to a network of workstations.

Bischof, c. H.↗

Hot Chips and Hot Interconnects for High End Computing Systems

I will discuss several processors: 1. The Cray proprietary processor used in the Cray X1; 2. The IBM Power 3 and Power 4 used in an IBM SP 3 and IBM SP 4 systems; 3. The Intel Itanium and Xeon, used in the SGI Altix systems and clusters respectively; 4. IBM System-on-a-Chip used in IBM BlueGene/L; 5. HP Alpha EV68 processor used in DOE ASCI Q cluster; 6. SPARC64 V processor, which is used in the Fujitsu PRIMEPOWER HPC2500; 7. An NEC proprietary processor, which is used in NEC SX-6/7; 8. Power 4+ processor, which is used in Hitachi SR11000; 9. NEC proprietary processor, which is used in Earth Simulator. The IBM POWER5 and Red Storm Computing Systems will also be discussed. The architectures of these processors will first be presented, followed by interconnection networks and a description of high-end computer systems based on these processors and networks. The performance of various hardware/programming model combinations will then be compared, based on latest NAS Parallel Benchmark results (MPI, OpenMP/HPF and hybrid (MPI + OpenMP). The tutorial will conclude with a discussion of general trends in the field of high performance computing, (quantum computing, DNA computing, cellular engineering, and neural networks).

Saini, Subhash↗

Research on Spectroscopy, Opacity, and Atmospheres

I propose to continue providing observers with basic data for interpreting spectra from stars, novas, supernovas, clusters, and galaxies. These data will include allowed and forbidden line lists, both laboratory and computed, for the first five to ten ions of all atoms and for all relevant diatomic molecules. I will eventually expand to all ions of the first thirty elements to treat far UV and X-ray spectra, and for envelope opacities. I also include triatomic molecules provided by other researchers. I have also made CDs with Partridge and Schwenke's water data for work on UV stars. The line data also serve as input to my model atmosphere and synthesis programs that generate energy distributions, photometry, limb darkening, and spectra that can be used for planning observations and for fitting observed spectra. The spectrum synthesis programs produce detailed plots with the lines identified. Grids of stellar spectra can be used for radial velocity-, rotation-, or abundance templates and for population synthesis. I am fitting spectra of bright stars to test the data and to produce atlases to guide observers. For each star the whole spectrum is computed from the UV to the far IR. The line data, opacities, models, spectra, and programs are freely distributed on CDs and on my Web site and represent a unique resource for many NASA programs. I am now in full production of new line lists for atoms. I am computing all ions of all elements from H to Zn and the first 5 ions of all the heavier elements, about 800 ions. For each ion I treat as many as 61 even and 61 odd configurations, computing all energy levels and eigenvectors. The Hamiltonian is determined from a scaled-Hartree-Fock starting guess by least squares fitting the observed energy levels. The average energy of each configuration is used in computing scaled-Thomas-Fermi-Dirac wavefunctions for each configuration which in turn are used to compute allowed and forbidden transition integrals. These are multiplied into the LS allowed and forbidden transition arrays. The transition arrays are transformed to the observed coupling to yield the allowed and forbidden line lists. Results are put on the web as they are finished. Provided I get funding,there will be more than 500 million lines. I will then compare ion by ion, to all the laboratory and computed data in the literature and make up a working line list for spectrum synthesis and opacity calculations with the best available data. As the laboratory spectrum analyses are improved, I will redo the calculations with the new energy levels. My original plan when I started the new calculations was to run through all the atoms using my old Cray programs from the 1980's that were limited to 1100 x 1100 arrays in the Hamiltonian for each J. Then I would go back and rerun the more complicated cases with 3000 x 3000 arrays so that I could include many more configurations and more configuration interactions. At present I am limited to 61 even and 61 odd configurations and I try to include everything up through n = 9. The current program runs on Alpha workstations. I decided to test the big program on Fe I and Fe II to see whether there was any great difference in the low configurations compared to those from the Cray program. Besides increasing the number of E1 lines by a factor of 6 to 7.7 million, there was an unexpected result: the electric quadrupole transitions were 10 times stronger than before because the transition integrals are weighted by r(exp 2) ---they become very large for high n, and because there are numerous configuration interactions that mix the low and high configurations. As a check I was able to reproduce Carstang's (1962) lower results by running his three configurations with my program. Since my model atom is still only a subset of a real Fe II ion, the true quadrupole A values are probably larger than mine. The magnetic dipole lines are affected by the mixing but the overall scale does not change. Because of this scovery I decided that there was no point in computing the small array cases. I have been running with as many configurations as I can and with thousands of parameters in the Hamiltonian. The computer runs take much longer to set up and produce than I had expected. I have concentrated on redoing the low iron group spectra, especially to get data for supernova modelers. I have done only Ca I -- Zn I, Ca II -- Zn II, CU I -- Cu XXIX, Zn I - Zn XXX, for practice at high stages of ionization, C I, C II, S I, and CL I and Ag I for people who were working on the laboratory spectra. Check my web site kurucz.harvard.edu for current additions. My latest calculations have been for carbon I and sulphur I, and silicon I is under way using the same elaborate approach as for C I, which took many months to do. These line lists greatly increase the number of lines in the ultraviolet, in the visible, and especially in the infrared. They will increase the opacity in A, F, and G stars. They will account for many unidentified lines in the sun.

Kurucz, Robert L.↗

Scalable in-memory RDFS closure on billions of triples.

We present an RDFS closure algorithm, specifically designed and implemented on the Cray XMT supercomputer, that obtains inference rates of 13 million inferences per second on the largest system configuration we used. The Cray XMT, with its large global memory (4TB for our experiments), permits the construction of a conceptually straightforward algorithm, fundamentally a series of operations on a shared hash table. Each thread is given a partition of triple data to process, a dedicated copy of the ontology to apply to the data, and a reference to the hash table into which it inserts inferred triples. The global nature of the hash table allows the algorithm to avoid a common obstacle for distributed memory machines: the creation of duplicate triples. On LUBM data sets ranging between 1.3 billion and 5.3 billion triples, we obtain nearly linear speedup except for two portions: file I/O, which can be ameliorated with the additional service nodes, and data structure initialization, which requires nearly constant time for runs involving 32 processors or more.

99 GENERAL AND MISCELLANEOUS↗

Co-design in ACES and exascale.

The Alliance for Computing at the Extreme Scale (ACES) is a Los Alamos and Sandia collaboration encompassing not only HPC procurements and operations, but also computer science and architecture research and development. One area of focus within ACES relates to the critical technology developments for future high performance computing systems and the applications that would run on them, and ACES is heavily involved in the proposed DOE Exascale Initiative. The proposed Exascale Initiative emphasizes the need for co-design, which is the three-way collaborative and concurrent design of HPC hardware, software, and the applications themselves. Transformational changes will occur not only in HPC hardware, but also in the applications space, and taken together these will require transformation changes in the overall software layers supporting the programming models, tools, runtimes, file systems, and operating systems. Co-design involving all three areas of hardware, software, and applications will be the key to success. This talk will outline key aspects of the Exascale Initiative and its emphasis on co-design. It will provide some examples from LANL & Sandia experiences in co-design including aspects of the innovative Roadrunner architecture and software, a ACES-Cray project studying advanced interconnects within Cray, and Sandia's work in computer system simulators and mini-applications.

99 GENERAL AND MISCELLANEOUS↗

Petascale Parallelization of the Gyrokinetic Toroidal Code

The Gyrokinetic Toroidal Code (GTC) is a global, three-dimensional particle-in-cell application developed to study microturbulence in tokamak fusion devices. The global capability of GTC is unique, allowing researchers to systematically analyze important dynamics such as turbulence spreading. In this work we examine a new radial domain decomposition approach to allow scalability onto the latest generation of petascale systems. Extensive performance evaluation is conducted on three high performance computing systems: the IBM BG/P, the Cray XT4, and an Intel Xeon Cluster. Overall results show that the radial decomposition approach dramatically increases scalability, while reducing the memory footprint - allowing for fusion device simulations at an unprecedented scale. After a decade where high-end computing (HEC) was dominated by the rapid pace of improvements to processor frequencies, the performance of next-generation supercomputers is increasingly differentiated by varying interconnect designs and levels of integration. Understanding the tradeoffs of these system designs is a key step towards making effective petascale computing a reality. In this work, we examine a new parallelization scheme for the Gyrokinetic Toroidal Code (GTC) [?] micro-turbulence fusion application. Extensive scalability results and analysis are presented on three HEC systems: the IBM BlueGene/P (BG/P) at Argonne National Laboratory, the Cray XT4 at Lawrence Berkeley National Laboratory, and an Intel Xeon cluster at Lawrence Livermore National Laboratory. Overall results indicate that the new radial decomposition approach successfully attains unprecedented scalability to 131,072 BG/P cores by overcoming the memory limitations of the previous approach. The new version is well suited to utilize emerging petascale resources to access new regimes of physical phenomena.

42 ENGINEERING↗

Lustre at Petascale: Experiences in Troubleshooting and Upgrading

Some veterans in the HPC industry semi-facetiously define supercomputers as devices that convert compute-bound problems into I/O-bound problems. Effective utilization of large high performance computing resources often requires access to large amounts of fast storage. The National Institute for Computational Sciences (NICS) operates Kraken, a 1.17 PetaFLOPS Cray XT5 for the National Science Foundation (NSF). Kraken s primary file system has migrated from Lustre 1.6 to 1.8 and is currently being moved to servers external to the machine. Additional bandwidth will be made available by mounting the NICS-wide Lustre file system. Newer versions of Lustre, beyond what Cray provides, are under evaluation for stability and performance. Over the past several years of operation, Kraken s Lustre file system has evolved to be extremely stable in an effort to better serve Kraken s users.

99 GENERAL AND MISCELLANEOUS↗

Data Movement Dominates: Final Report

Over the past three years in this project, what we have observed is that the primary reason for data movement in large-scale systems is that the per-node capacity is not large enough—i.e., one of the solutions to the data-movement problem (certainly not the only solution that is required, but a significant one nonetheless) is to increase per-node capacity so that inter-node traffic is reduced. This unfortunately is not as simple as it sounds. Today’s main memory systems for datacenters, enterprise computing systems, and supercomputers, fail to provide high per-socket capacity [Dirik & Jacob 2009; Cooper-Balis et al. 2012], except at extremely high price points (factors of 10–100x the cost/bit of consumer main-memory systems) [Stokes 2008]. The reason is that our choice of technology for today’s main memory systems—i.e., DRAM, which we have used as a main-memory technology since the 1970s [Jacob et al. 2007]—can no longer keep up with our needs for density and price per bit. Main memory systems have always been built from the cheapest, densest, lowest-power memory technology available, and DRAM is no longer the cheapest, the densest, nor the lowest-power storage technology out there. It is now time for DRAM to go the way that SRAM went: move out of the way for a cheaper, slower, denser storage technology, and become a cache instead. This inflection point has happened before, in the context of SRAM yielding to DRAM. There was once a time that SRAM was the storage technology of choice for all main memories [Tomasulo 1967; Thornton 1970; Kidder 1981]. However, once DRAM hit volume production in the 1970s and 80s, it supplanted SRAM as a main memory technology because it was cheaper, and it was denser. It also happened to be lower power, but that was not the primary consideration of the day. At the time, it was recognized that DRAM was much slower than SRAM, but it was only at the supercomputer level (For instance the Cray X-MP in the 1980s and its follow-on, the Cray Y-MP, in the 1990s) that could one afford to build ever- larger main memories out of SRAM—the reasoning for moving to DRAM was that an appropriately designed memory hierarchy, built of DRAM as main memory and SRAM as a cache, would approach the performance of SRAM, at the price-per-bit of DRAM [Mashey 1999]. Today it is quite clear that, were one to build an entire multi-gigabyte main memory out of SRAM instead of DRAM, one could improve the performance of almost any computer system by up to an order of magnitude—but this option is not even considered, because to build that system would be prohibitively expensive. It is now time to revisit the same design choice in the context of modern technologies and modern systems. For reasons both technical and economic, we can no longer afford to build ever-larger main memory systems out of DRAM. Flash memory, on the other hand, is significantly cheaper and denser than DRAM and therefore should take its place. While it is true that flash is significantly slower than DRAM, one can afford to build much larger main memories out of flash than out of DRAM, and we show that an appropriately designed memory hierarchy, built of flash as main memory and DRAM as a cache, will approach the performance of DRAM, at the price-per-bit of flash. In our studies as part of this project, we have investigated Non-Volatile Main Memory (NVMM), a new main-memory architecture for large-scale computing systems, one that is specifically designed to address the weaknesses described previously. In particular, it provides the following features: non-volatility: The bulk of the storage is comprised of NAND flash, and in this organization DRAM is used only as a cache, not as main memory. Furthermore, the flash is journaled, which means that operations such as checkpoint/restore are already built into the system. 1+ terabytes of storage per socket: SSDs and DRAM DIMMs have roughly the same form factor (several square inches of PCB surface area), and terabyte SSDs are now commonplace. performance approaching that of DRAM: DRAM is used as a cache to the flash system. price-per-bit approaching that of NAND: Flash is currently well under $0.50 per gigabyte; DDR3 SDRAM is currently just over $10 per gigabyte [Newegg 2014]. Even today, one can build an easily affordable main memory system with a terabyte or more of NAND storage per CPU socket (which would be extremely expensive were one to use DRAM), and our cycle- accurate, full-system experiments show that this can be done at a performance point that lies within a factor of two of DRAM.

96 KNOWLEDGE MANAGEMENT AND PRESERVATION↗

OpenSHMEM-UCX : Evaluation of UCX for implementing OpenSHMEM Programming Model

The OpenSHMEM reference implementation was developed towards the goal of developing an open source and high-performing Open- SHMEM implementation. To achieve portability and performance across various networks, the OpenSHMEM reference implementation uses GAS- Net and UCCS for network operations. Recently, new network layers have emerged with the promise of providing high-performance, scalabil- ity, and portability for HPC applications. In this paper, we implement the OpenSHMEM reference implementation to use the UCX framework for network operations. Then, we evaluate its performance and scalabil- ity on Cray XK systems to understand UCX s suitability for developing the OpenSHMEM programming model. Further, we develop a bench- mark called SHOMS for evaluating the OpenSHMEM implementation. Our experimental results show that OpenSHMEM-UCX outperforms the vendor supplied OpenSHMEM implementation in most cases on the Cray XK system by up to 40% with respect to message rate and up to 70% for the execution of application kernels.

Baker, Matthew B↗

Tuning HDF5 subfiling performance on parallel file systems

Subfiling is a technique used on parallel file systems to reduce locking and contention issues when multiple compute nodes interact with the same storage target node. Subfiling provides a compromise between the single shared file approach that instigates the lock contention problems on parallel file systems and having one file per process, which results in generating a massive and unmanageable number of files. In this paper, we evaluate and tune the performance of recently implemented subfiling feature in HDF5. In specific, we explain the implementation strategy of subfiling feature in HDF5, provide examples of using the feature, and evaluate and tune parallel I/O performance of this feature with parallel file systems of the Cray XC40 system at NERSC (Cori) that include a burst buffer storage and a Lustre disk-based storage. We also evaluate I/O performance on the Cray XC30 system, Edison, at NERSC. Our results show performance benefits of 1.2X to 6X performance advantage with subfiling compared to writing a single shared HDF5 file. We present our exploration of configurations, such as the number of subfiles and the number of Lustre storage targets to storing files, as optimization parameters to obtain superior I/O performance. Based on this exploration, we discuss recommendations for achieving good I/O performance as well as limitations with using the subfiling feature.

97 MATHEMATICS AND COMPUTING↗

TOKIO on ClusterStor: Connecting Standard Tools to Enable Holistic I/O Performance Analysis

At present, I/O performance analysis requires different tools to characterize individual components of the I/O subsystem, and institutional I/O expertise is relied upon to translate these disparate data into an integrated view of application performance. This process is labor-intensive and not sustainable as the storage hierarchy deepens and system complexity increases. To address this growing disparity, we have developed the Total Knowledge of I/O (TOKIO) framework to combine the insights from existing component-level monitoring tools and provide a holistic view of performance across the entire I/O stack. A reference implementation of TOKIO, pytokio, is presented here. Using monitoring tools included with Cray XC and ClusterStor systems alongside commonly deployed community-supported tools, we demonstrate how pytokio provides a lightweight foundation for holistic I/O performance analyses on two Cray XC systems deployed at different HPC centers. We present results from integrated analyses that allow users to quantify the degree of I/O contention that affected their jobs and probabilistically identify unhealthy storage devices that impacted their performance.We also apply pytokio to inspect the utilization of NERSC’s DataWarp burst buffer and demonstrate how pytokio can be used to identify users and applications who may stand to benefit most from migrating their workloads from Lustre to the burst buffer.

Lockwood, Glenn K.↗

OpenMP and MPI implementations of an elasto-viscoplastic fast Fourier transform-based micromechanical solver for fast crystal plasticity modeling

Here, we explore several parallel implementations of an elasto-viscoplastic fast Fourier transform (EVPFFT) model using Message Passing Interface (MPI), OpenMP, and a hybrid of MPI and OpenMP to efficiently predict micromechanical response of polycrystals. Performance studies using EVPFFT are performed based on domain decomposition over voxels of a periodic cell, which is a representative volume element (RVE) of polycrystalline copper. We begin by parallelizing the computationally intensive Newton–Raphson (NR) single crystal solver within EVPFFT. Next, we compare the performance of the serial and parallel FFTW (Fastest Fourier Transform in the West) using OpenMP (OpenMP-FFTW) and MPI (MPI-FFTW) with the original Numerical Recipes-based FOURN routine within EVPFFT. In the parallel environment, we find that the FFT calculations are best performed using the MPI version of FFTW. Finally, the remainder of the code, except read/write subroutines, is parallelized. Significant speedups of the original EVPFFT model are achieved using MPI on shared memory multicore workstations. Furthermore, results achieved on a distributed memory Cray supercomputer show promising strong and weak scalability and in some cases even super scalability for the single crystal NR solver in EVPFFT. MPI-FFTW also scales perfectly for microstructure RVEs larger than 64 3 FFT voxels. For example, the MPI-EVPFFT parallel version of the code accelerates the simulations for approximately two orders of magnitude using 64 cores over the old serial code for an RVE size of 128 3 . Finally, the parallel EVPFFT code developed in this work can run massive voxel-based microstructural RVEs taking the advantages of thousands of logical cores provided by more advanced clusters.

42 ENGINEERING↗

A High-Performance Hybrid Computing Approach to Massive Contingency Analysis in the Power Grid

Operating the electrical power grid to prevent power black-outs is a complex task. An important aspect of this is contingency analysis, which involves understanding and mitigating potential failures in power grid elements such as transmission lines. When taking into account the potential for multiple simultaneous failures (known as the N-x contingency problem), contingency analysis becomes a massively computational task. In this paper we describe a novel hybrid computational approach to contingency analysis. This approach exploits the unique graph processing performance of the Cray XMT in conjunction with a conventional massively parallel compute cluster to identify likely simultaneous failures that could cause widespread cascading power failures that have massive economic and social impact on society. The approach has the potential to provide the first practical and scalable solution to the N-x contingency problem. When deployed in power grid operations, it will increase the grid operator’s ability to deal effectively with outages and failures with power grid components while preserving stable and safe operation of the grid. The paper describes the architecture of our solution and presents preliminary performance results that validate the efficacy of our approach.

24 POWER TRANSMISSION AND DISTRIBUTION↗

Massive Social Network Analysis: Mining Twitter for Social Good

Social networks produce an enormous quantity of data. Facebook consists of over 400 million active users sharing over 5 billion pieces of information each month. Analyzing this vast quantity of unstructured data presents challenges for software and hardware. We present GraphCT, a Graph Characterization Tooklit for massive graphs representing social network data. On a 128-processor Cray XMT, GraphCT estimates the betweenness centrality of an artificially generated (R-MAT) 537 million vertex, 8.6 billion edge graph in 55 minutes. We use GraphCT to analyze public data from Twitter, a microblogging network. Twitter's message connections appear primarily tree-structured as a news dissemination system. Within the public data, however, are clusters of conversations. Using GraphCT, we can rank actors within these conversations and help analysts focus attention on a much smaller data subset.

99 GENERAL AND MISCELLANEOUS↗