High Performance Computing Approach Research Articles

A highly efficient computational approach for fast scan-resolved simulations of metal additive manufacturing processes on the scale of real parts

This article proposes a novel high-performance computing approach for the prediction of the temperature field in powder bed fusion (PBF) additive manufacturing (AM) processes. In contrast to many existing approaches to part-scale simulations, the underlying computational model consistently resolves physical scan tracks without additional heat source scaling, agglomeration strategies or any other heuristic modeling assumptions. A growing, adaptively refined mesh accurately captures all details of the laser beam motion. Critically, the fine spatial resolution required for resolved scan tracks in combination with the high scan velocities underlying these processes mandates the use of comparatively small time steps to resolve the underlying physics. Explicit time integration schemes are well-suited for this setting, while unconditionally stable implicit time integration schemes are employed for the interlayer cool down phase governed by significantly larger time scales. These two schemes are combined and implemented in an efficient fast operator evaluation framework providing significant performance gains and optimization opportunities. The capabilities of the novel framework are demonstrated through realistic AM examples on the centimeter scale including the first scan-resolved simulation of the entire NIST AM Benchmark cantilever specimen, with a computation time of less than one day. Apart from physical insights gained through these simulation examples, also numerical aspects are thoroughly studied on basis of weak and strong parallel scaling tests. As potential applications, the proposed thermal PBF simulation approach can serve as a basis for microstructure and thermo-mechanical predictions on the part-scale, but also to assess the influence of scan pattern and part geometry on melt pool shape and temperature, which are important indicators for well-known process instabilities.

High-Performance and Parallel Computing Techniques Review: Applications, Challenges and Potentials to Support Net-Zero Transition of Future Grids

The transition towards net-zero emissions is inevitable for humanity’s future. Of all the sectors, electrical energy systems emit the most emissions. This urgently requires the witnessed accelerating technological landscape to transition towards an emission-free smart grid. It involves massive integration of intermittent wind and solar-powered resources into future power grids. Additionally, new paradigms such as large-scale integration of distributed resources into the grid, proliferation of Internet of Things (IoT) technologies, and electrification of different sectors are envisioned as essential enablers for a net-zero future. However, these changes will lead to unprecedented size, complexity and data of the planning and operation problems of future grids. It is thus important to discuss and consider High Performance Computing (HPC), parallel computing, and cloud computing prospects in any future electrical energy studies. This article recounts the dawn of parallel computation in power system studies, providing a thorough history and paradigm background for the reader, leading to the most impactful recent contributions. The reviews are split into Central Processing Unit (CPU) based, Graphical Processing Unit (GPU) based, and Cloud-based studies and smart grid applications. The state-of-the-art is also discussed, highlighting the issue of standardization and the future of the field. The reviewed papers are predominantly focused on classical imperishable electrical system problems. This indicates the need for further research on parallel and HPC approaches applied to future smarter grid challenges, particularly to the integration of renewable energy into the smart grid.

HPC-enabling technologies for high-fidelity combustion simulations

With the increase in computational power in the last decade and the forthcoming Exascale supercomputers, a new horizon in computational modelling and simulation is envisioned in combustion science. Considering the multiscale and multiphysics characteristics of turbulent reacting flows, combustion simulations are considered as one of the most computationally demanding applications running on cutting-edge supercomputers. Exascale computing opens new frontiers for the simulation of combustion systems as more realistic conditions can be achieved with high-fidelity methods. However, an efficient use of these computing architectures requires methodologies that can exploit all levels of parallelism. The efficient utilization of the next generation of supercomputers needs to be considered from a global perspective, that is, involving physical modelling and numerical methods with methodologies based on High-Performance Computing (HPC) and hardware architectures. This review introduces recent developments in numerical methods for large-eddy simulations (LES) and direct-numerical simulations (DNS) to simulate combustion systems, with focus on the computational performance and algorithmic capabilities. Due to the broad scope, a first section is devoted to describe the fundamentals of turbulent combustion, which is followed by a general description of state-of-the-art computational strategies for solving these problems. These applications require advanced HPC approaches to exploit modern supercomputers, which is addressed in the third section. The increasing complexity of new computing architectures, with tightly coupled CPUs and GPUs, as well as high levels of parallelism, requires new parallel models and algorithms exposing the required level of concurrency. Advances in terms of dynamic load balancing, vectorization, GPU acceleration and mesh adaptation have permitted to achieve highly-efficient combustion simulations with data-driven methods in HPC environments. Therefore, dedicated sections covering the use of high-order methods for reacting flows, integration of detailed chemistry and two-phase flows are addressed. Final remarks and directions of future work are given at the end.

Real-time acoustic modeling in the 2021 New England Shelf Break Acoustics experiment and beyond

Real-time acoustic modeling in a shipboard context emphasizes the element of computation time in pursuit of in situ comparison between observed ocean data and model output. The importance of computation time highlights two competing factors in efforts to produce real-time acoustic predictions: the quantity of models to run and model accuracy. During the 2021 New England Shelf Break Acoustics (NESBA) experiment, a four-dimensional (4D) model of acoustic signals in the experiment network comprised of multiple stationary sound sources and receivers was produced using shipboard high-performance computing (HPC) resources. This effort involved hundreds of acoustic model runs across four separate source and receiver pairings and allowed for in-situ model adjustments, ultimately producing favorable comparisons between data and models in a matter of days. This presentation will examine the HPC approaches taken during the NESBA science cruise and address the planned improvements to the system for future scientific field experiments. [This work is supported by the Office of Naval Research.]

An HPC hybrid parallel approach to the experimental analysis of Fermat’s theorem extension to arbitrary dimensions on heterogeneous computer systems

In this work, we consider the empirical, computational, study of the generalized Fermat’s last theorem conjecture recently proposed using Minkowski norms defined within natural vector semispaces. This kind of studies relies on the connection of vector spaces to Boolean hypercubes, which relates to systematic natural number generation and, thus, to the inner structure of natural vector sets. This topic is of interest for the vector description of molecular structures among other uses associated with the development of new Quantitative Structure-Properties Relationships, QSPR. Here, we transform the problem into a combinatorial one susceptible of computational analysis. Due to the computational demands of these studies, we resort to a High-Performance Computing (HPC) approach. Thus, arbitrary precision arithmetic is introduced as well as new algorithmic approaches for the computation of Minkowski norms and the application of Fermat’s condition. Besides, the problem is transformed into an embarrassingly parallel one. This permits to develop a hybrid shared-distributed memory parallel approach. To optimize the efficiency of the process on heterogeneous computer systems, a hierarchical dynamic self-scheduling approach is introduced. We find an increase in efficiency of 23% when using the proposed self-scheduling strategy. Application of this HPC approach to about 6000 billion three-dimensional natural vectors, for Boolean hypercubes up to dimension 15, and a range of Minkowski norm powers between 3 and 64, strongly suggests conjecturing that Fermat’s last theorem can be generalized from dimension two to dimension three.

Accelerating a Geometrical Approximated PCA Algorithm Using AVX2 and CUDA

Remote sensing data has known an explosive growth in the past decade. This has led to the need for efficient dimensionality reduction techniques, mathematical procedures that transform the high-dimensional data into a meaningful, reduced representation. Projection Pursuit (PP) based algorithms were shown to be efficient solutions for performing dimensionality reduction on large datasets by searching low-dimensional projections of the data where meaningful structures are exposed. However, PP faces computational difficulties in dealing with very large datasets—which are common in hyperspectral imaging, thus raising the challenge for implementing such algorithms using the latest High Performance Computing approaches. In this paper, a PP-based geometrical approximated Principal Component Analysis algorithm (gaPCA) for hyperspectral image analysis is implemented and assessed on multi-core Central Processing Units (CPUs), Graphics Processing Units (GPUs) and multi-core CPUs using Single Instruction, Multiple Data (SIMD) AVX2 (Advanced Vector eXtensions) intrinsics, which provide significant improvements in performance and energy usage over the single-core implementation. Thus, this paper presents a cross-platform and cross-language perspective, having several implementations of the gaPCA algorithm in Matlab, Python, C++ and GPU implementations based on NVIDIA Compute Unified Device Architecture (CUDA). The evaluation of the proposed solutions is performed with respect to the execution time and energy consumption. The experimental evaluation has shown not only the advantage of using CUDA programming in implementing the gaPCA algorithm on a GPU in terms of performance and energy consumption, but also significant benefits in implementing it on the multi-core CPU using AVX2 intrinsics.

A Computational Pipeline for the Extraction of Actionable Biological Information From NGS-Phage Display Experiments.

Phage Display is a powerful method for the identification of peptide binding to targets of variable complexities and tissues, from unique molecules to the internal surfaces of vessels of living organisms. Particularly for in vivo screenings, the resulting repertoires can be very complex and difficult to study with traditional approaches. Next Generation Sequencing (NGS) opened the possibility to acquire high resolution overviews of such repertoires and thus facilitates the identification of binders of interest. Additionally, the ever-increasing amount of available genome/proteome information became satisfactory regarding the identification of putative mimicked proteins, due to the large scale on which partial sequence homology is assessed. However, the subsequent production of massive data stresses the need for high-performance computational approaches in order to perform standardized and insightful molecular network analysis. Systems-level analysis is essential for efficient resolution of the underlying molecular complexity and the extraction of actionable interpretation, in terms of systemic biological processes and pathways that are systematically perturbed. In this work we introduce PepSimili, an integrated workflow tool, which performs mapping of massive peptide repertoires on whole proteomes and delivers a streamlined, systems-level biological interpretation. The tool employs modules for modeling and filtering of background noise due to random mappings and amplifies the biologically meaningful signal through coupling with BioInfoMiner, a systems interpretation tool that employs graph-theoretic methods for prioritization of systemic processes and corresponding driver genes. The current implementation exploits the Galaxy environment and is available online. A case study using public data is presented, with and without a control selection.

Hyperparameter optimization of neural network-driven spatial models accelerated using cyber-enabled high-performance computing

ABSTRACTArtificial neural networks (ANNs) have been extensively used for the spatially explicit modeling of complex geographic phenomena. However, because of the complexity of the computational process, there has been an inadequate investigation on the parameter configuration of neural networks. Most studies in the literature from GIScience rely on a trial-and-error approach to select the parameter setting for ANN-driven spatial models. Hyperparameter optimization provides support for selecting the optimal architectures of ANNs. Thus, in this study, we develop an automated hyperparameter selection approach to identify optimal neural networks for spatial modeling. Further, the use of hyperparameter optimization is challenging because hyperparameter space is often large and the associated computational demand is heavy. Therefore, we utilize high-performance computing to accelerate the model selection process. Furthermore, we involve spatial statistics approaches to improve the efficiency of hyperparameter optimization. The spatial model used in our case study is a land price evaluation model in Mecklenburg County, North Carolina, USA. Our results demonstrate that the automated selection approach improves the model-level performance compared with linear regression, and the high-performance computing and spatial statistics approaches are of great help for accelerating and enhancing the selection of optimal neural networks for spatial modeling.

Performance and energy-efficient implementation of a smart city application on FPGAs

The continuous growth of modern cities and the request for better quality of life, coupled with the increased availability of computing resources, lead to an increased attention to smart city services. Smart cities promise to deliver a better life to their inhabitants while simultaneously reducing resource requirements and pollution. They are thus perceived as a key enabler to sustainable growth. Out of many other issues, one of the major concerns for most cities in the world is traffic, which leads to a huge waste of time and energy, and to increased pollution. To optimize traffic in cities, one of the first steps is to get accurate information in real time about the traffic flows in the city. This can be achieved through the application of automated video analytics to the video streams provided by a set of cameras distributed throughout the city. Image sequence processing can be performed both peripherally and centrally. In this paper, we argue that, since centralized processing has several advantages in terms of availability, maintainability and cost, it is a very promising strategy to enable effective traffic management even in large cities. However, the computational costs are enormous, and thus require an energy-efficient High-Performance Computing approach. Field Programmable Gate Arrays (FPGAs) provide comparable computational resources to CPUs and GPUs, yet require much lower amounts of energy per operation (around 6times and 10times for the application considered in this case study). They are thus preferred resources to reduce both energy supply and cooling costs in the huge datacenters that will be needed by Smart Cities. In this paper, we describe efficient implementations of high-performance algorithms that can process traffic camera image sequences to provide traffic flow information in real-time at a low energy and power cost.

Applying big data paradigms to a large scale scientific workflow: Lessons learned and future directions

The increasing amounts of data related to the execution of scientific workflows has raised awareness of their shift towards parallel data-intensive problems. In this paper, we deliver our experience combining the traditional high-performance computing and grid-based approaches with Big Data analytics paradigms, in the context of scientific ensemble workflows. Our goal was to assess and discuss the suitability of such data-oriented mechanisms for production-ready workflows, especially in terms of scalability. We focused on two key elements in the Big Data ecosystem: the data-centric programming model, and the underlying infrastructure that integrates storage and computation in each node. We experimented with a representative MPI-based iterative workflow from the hydrology domain, EnKF-HGS, which we re-implemented using the Spark data analysis framework. We conducted experiments on a local cluster, a private cloud running OpenNebula, and the Amazon Elastic Compute Cloud (AmazonEC2). The results we obtained were analysed to synthesize the lessons we learned from this experience, while discussing promising directions for further research.

Mechanical Characterization of the Vessel Wall by Data Assimilation of Intravascular Ultrasound Studies

Atherosclerotic plaque rupture and erosion are the most important mechanisms underlying the sudden plaque growth, responsible for acute coronary syndromes and even fatal cardiac events. Advances in the understanding of the culprit plaque structure and composition are already reported in the literature, however, there is still much work to be done toward in-vivo plaque visualization and mechanical characterization to assess plaque stability, patient risk, diagnosis and treatment prognosis. In this work, a methodology for the mechanical characterization of the vessel wall plaque and tissues is proposed based on the combination of intravascular ultrasound (IVUS) imaging processing, data assimilation and continuum mechanics models within a high performance computing (HPC) environment. Initially, the IVUS study is gated to obtain volumes of image sequences corresponding to the vessel of interest at different cardiac phases. These sequences are registered against the sequence of the end-diastolic phase to remove transversal and longitudinal rigid motions prescribed by the moving environment due to the heartbeat. Then, optical flow between the image sequences is computed to obtain the displacement fields of the vessel (each associated to a certain pressure level). The obtained displacement fields are regarded as observations within a data assimilation paradigm, which aims to estimate the material parameters of the tissues within the vessel wall. Specifically, a reduced order unscented Kalman filter is employed, endowed with a forward operator which amounts to address the solution of a hyperelastic solid mechanics model in the finite strain regime taking into account the axially stretched state of the vessel, as well as the effect of internal and external forces acting on the arterial wall. Due to the computational burden, a HPC approach is mandatory. Hence, the data assimilation and computational solid mechanics computations are parallelized at three levels: (i) a Kalman filter level; (ii) a cardiac phase level; and (iii) a mesh partitioning level. To illustrate the capabilities of this novel methodology toward the in-vivo analysis of patient-specific vessel constituents, mechanical material parameters are estimated using in-silico and in-vivo data retrieved from IVUS studies. Limitations and potentials of this approach are exposed and discussed.

High Performance Computing for Regional Building Seismic Damage Simulation

Modern cities are becoming integrated systems that consist of a high density of population and buildings. Once they are hit by earthquakes, the damage or collapse of buildings will result in huge economic losses and casualties. For the three key problems (i.e. modeling, computing and visualization) in regional building seismic damage simulation, a high-fidelity modeling method, a high-performance computing approach and a realistic visualization technique are proposed in this study, respectively. With respect to the modeling problem, a multi-degree-of-freedom (MDOF) shear model for multi-story buildings and a MDOF flexural-shear model for tall buildings are proposed. Also, the corresponding parameter determination methods are proposed. For the computing problem, a parallel computing method based on GPU (graphics processing unit) for regional building seismic damage simulation is proposed, leading to a speedup ratio of approximately 54 times. In regard to the visualization problem, a 3D-GIS data generation method using the widely accessible 3D urban polygonal model and a realistic visualization method are proposed for displaying the simulated urban earthquake disaster scenario. The above proposed methods have been used in a large-sized city and Beijing central business district (CBD). Accurate, efficient, and realistic simulations for regional building seismic damage have been achieved with these methods. The outcome of this study provides important technology for regional building seismic damage simulation, which can improve the ability of urban disaster prevention and mitigation.

High performance computing in modelling and simulation

With the arrival of high performance computing (HPC) technologies, one of the most well-known scientific fields that succeeded in taking advantage of the new skyrocketing HPC opportunities was the modelling and simulation of computationally intensive scientific applications. It is now clear that the development of such models through which computers can extensively simulate the evolution of artificial and natural systems is fundamental for the advancement of science. As a result, different HPC approaches have allowed researchers nowadays to considerably extend the application of computing methodologies in research and industry, but also extend it to the quantitative study of complex phenomena. Moreover, this has permitted a broader and more effective application of numerical methods for differential equation systems (e.g. finite element method (FEM), finite difference method (FDM), etc.) on the one hand, and, most importantly, the application of alternative computational paradigms, such as cellular automata (CA), genetic algorithms, neural networks, swarm intelligence, and so on, on the other. The latter have demonstrated their effectiveness for modelling purposes when traditional simulation methodologies have proven to be impracticable. In conjunction with high parallel computer architectures, like the ones found in graphics processing units (GPUs), with thousands of computational threads on each of them, they are applied as fine candidates to overcome the computational burden imposed by vastly different scientific applications in diverse fields, such as engineering, physics, chemistry, biology, geology, medicine, ecology, sociology, traffic control, economy, and so on. Nevertheless, the resulting new parallelization paradigms and novel parallel computational approaches for all the aforementioned applications are considered one of the hot topics of today’s research. Having all this in mind, this special issue aims to provide a platform for a multidisciplinary community composed of scholars, researchers, developers, educators, practitioners and experts from world-leading universities, institutions, agencies and companies in computational science, and thus in the high performance computing for modelling and simulation field. As a sequel to the special session on high performance computing in modelling and simulation (HPCMS) within the 22nd Euromicro international conference on parallel, distributed and network-based processing (PDP), held in Turin, Italy on 12–14 February 2014, our goal in organizing this special issue of the ‘International Journal of High Performance Computing Applications’ was to gather the most suitable publications advancing HPCMS from both a theoretical and an application point of view and which were presented in an earlier reduced version in PDP 2014. We are pleased to say that this goal was achieved by the four excellent contributions presented in this issue, briefly summarized in the following. More specifically, this special issue addresses inherent parallel models and complex numerical models and their implementations by using GPUs as a fundamental computational primitive in various application fields like electromagnetics, environmental modelling and aircraft evacuation. Moreover, the evaluation of a one-instruction-per-cycle (one-IPC) high-level simulator of the microthreaded many-core systems is provided in detail while integrating the architecture model with the application model by abstracting the details of instruction execution. The first paper, by Camargos et al., provides a performance analysis of a parallel implementation for both preconditioned conjugate gradient and preconditioned bi-conjugate gradient solvers running on GPUs with a compute unified device architecture (CUDA) programming model. As a result, they focused on a GPU-accelerated iterative solution of complex-entry systems issued from 3D edge-Finite Element Analysis (FEA) of electromagnetics in the frequency domain. Their results

Circulant weighing matrices: a demanding challenge for parallel optimization metaheuristics

Circulant weighing matrices constitute a special type of combinatorial matrices that have attracted scientific interest for many years. The existence and determination of specific classes of circulant weighing matrices remains an active research area that involves both theoretical algebraic techniques as well as high-performance computational optimization approaches. The present work aims at investigating the potential of four established parallel metaheuristics as well as a special Algorithm Portfolio approach, on solving such problems. For this purpose, the algorithms are applied on a hard circulant weighing matrix existence problem. The obtained results are promising, offering insightful conclusions.

Genomic Resources for Water Yam (Dioscorea alata L.): Analyses of EST-Sequences, De Novo Sequencing and GBS Libraries.

The reducing cost and rapid progress in next-generation sequencing techniques coupled with high performance computational approaches have resulted in large-scale discovery of advanced genomic resources in several model and non-model plant species. Yam (Dioscorea spp.) is a major food and cash crop in many countries but research efforts have been limited to understand the genetics and generate genomic information for the crop. The availability of a large number of genomic resources including genome-wide molecular markers will accelerate the breeding efforts and application of genomic selection in yams. In the present study, several methods including expressed sequence tags (EST)-sequencing, de novo sequencing, and genotyping-by-sequencing (GBS) profiles on two yam (Dioscorea alata L.) genotypes (TDa 95/00328 and TDa 95-310) was performed to generate genomic resources for use in its improvement programs. This includes a comprehensive set of EST-SSRs, genomic SSRs, whole genome SNPs, and reduced representation SNPs. A total of 1,152 EST-SSRs were developed from >40,000 EST-sequences generated from the two genotypes. A set of 388 EST-SSRs were validated as polymorphic showing a polymorphism rate of 34% when tested on two diverse parents targeted for anthracnose disease. In addition, approximately 40X de novo whole genome sequence coverage was generated for each of the two genotypes, and a total of 18,584 and 15,952 genomic SSRs were identified for TDa 95/00328 and TDa 95-310, respectively. A custom made pipeline resulted in the selection of 573 genomic SSRs common across the two genotypes, of which only eight failed, 478 being polymorphic and 62 monomorphic indicating a polymorphic rate of 83.5%. Additionally, 288,505 high quality SNPs were also identified between these two genotypes. Genotyping by sequencing reads on these two genotypes also revealed 36,790 overlapping SNP positions that are distributed throughout the genome. Our efforts in using different approaches in generating genomic resources provides a non-biased glimpse into the publicly available EST-sequences, yam genome, and GBS profiles with affirmation that the genomic complexity can be methodically unraveled and constitute a critical foundation for future studies in linkage mapping, germplasm analysis, and predictive breeding.

Dual-scale phase-field simulation of Mg-Al alloy solidification

Phase-field simulations of the nucleation and growth of primary α-Mg phase as well as secondary, β-phase of a Mg-Al alloy are presented. The nucleation model for α- and β-Mg phases is based on the “free growth model” by Greer et al.. After the α-Mg phase solidification we study a divorced eutectic growth of α- and β-Mg phases in a zoomed in melt channel between α-phase dendrites. The simulated cooling curves and final microstructures of α-grains are compared with experiments. In order to further enhance the resolution of the interdendritic region a high-performance computing approach has been used allowing significant simulation speed gain when using supercomputing facilities.

A High Performance Computing Approach to the Simulation of Fluid-Solid interaction Problems with Rigid and Flexible Components

Abstract This work outlines a unified multi-threaded, multi-scale High Performance Computing (HPC) approach for the direct numerical simulation of Fluid-Solid Interaction (FSI) problems. The simulation algorithm relies on the extended Smoothed Particle Hydrodynamics (XSPH) method, which approaches the fluid flow in a La-grangian framework consistent with the Lagrangian tracking of the solid phase. A general 3D rigid body dynamics and an Absolute Nodal Coordinate Formulation (ANCF) are implemented to model rigid and flexible multibody dynamics. The two-way coupling of the fluid and solid phases is supported through use of Boundary Condition Enforcing (BCE) markers that capture the fluid-solid coupling forces by enforcing a no-slip boundary condition. The solid-solid short range interaction, which has a crucial impact on the small-scale behavior of fluid-solid mixtures, is resolved via a lubrication force model. The collective system states are integrated in time using an explicit, multi-rate scheme. To alleviate the heavy computational load, the overall algorithm leverages parallel computing on Graphics Processing Unit (GPU) cards. Performance and scaling analysis are provided for simulations scenarios involving one or multiple phases with up to tens of thousands of solid objects. The software implementation of the approach, called Chrono:Fluid, is part of the Chrono project and available as an open-source software.

DAIRRy-BLUP: A High-Performance Computing Approach to Genomic Prediction

In genomic prediction, common analysis methods rely on a linear mixed-model framework to estimate SNP marker effects and breeding values of animals or plants. Ridge regression-best linear unbiased prediction (RR-BLUP) is based on the assumptions that SNP marker effects are normally distributed, are uncorrelated, and have equal variances. We propose DAIRRy-BLUP, a parallel, Distributed-memory RR-BLUP implementation, based on single-trait observations ( Y: ), that uses the Average Information algorithm for restricted maximum-likelihood estimation of the variance components. The goal of DAIRRy-BLUP is to enable the analysis of large-scale data sets to provide more accurate estimates of marker effects and breeding values. A distributed-memory framework is required since the dimensionality of the problem, determined by the number of SNP markers, can become too large to be analyzed by a single computing node. Initial results show that DAIRRy-BLUP enables the analysis of very large-scale data sets (up to 1,000,000 individuals and 360,000 SNPs) and indicate that increasing the number of phenotypic and genotypic records has a more significant effect on the prediction accuracy than increasing the density of SNP arrays.

Open Source Software to Visualize Complex Data on Remote CEA's Supercomputing Facilities

In order to guaranty performances of complex systems using numerical simulation, CEA is performing advanced data analysis and scientific visualization with open source software using High Performance Computing (HPC) capability. The diversity of the physics to study produces results of growing complexity in terms of large-scale, high dimensional and multivariate data. Moreover, the HPC approach introduces another layer of complexity by allowing computation amongst thousands of remote cores accessed from sites located hundreds of kilometers away from the computing facility. This paper presents how CEA deploys and contributes to open source software to enable production class visualization tools in a high performance computing context. Among several open source projects used at CEA, this presentation will focus on Visit, VTK and Paraview. In the first part we will address specific issues encountered when deploying VisIt and Paraview in a multi-site supercomputing facility for end-users. Several examples will be given on how such tools can be adapted to take advantage of a parallel setting to explore large multi-block dataset or perform remote visualization on material interface reconstructions of billions of cells. Then, the specific challenges faced to deliver Paraview’s Catalyst capabilities to end-users will be discussed. In the second part, we will describe how CEA contributes to open source visualization software and associated software development strategy by emphasizing on two recent development projects. The first is an integrated simulation workbench providing plugins for every step required to achieve numerical simulation independently on a local or a remote computer. Embedded in an Eclipse RCP environment, VTK views allow the users to perform data input using interaction or mesh preview before running the simulation code. Contributions to VTK have been made in order to smoothly integrate these technologies. The second details how recent developments at CEA have helped to visualize and to analyze results from ExaStamp, a parallel molecular dynamics simulation code dealing with molecular systems ranging from a few millions up to a billion atoms. These developments include a GPU intensive rendering method specialized for atoms and specific parallel algorithms to process molecular data sets.

Cost-oriented proactive fault tolerance approach to high performance computing (HPC) in the cloud

Cloud computing offers new computing paradigms, capacity and flexible solutions to high performance computing (HPC) applications. For example, Hardware as a Service (HaaS) allows users to provide a large number of virtual machines (VMs) for computation-intensive applications using the HaaS model. Due to the large number of VMs and electronic components in HPC system in the cloud, any fault during the execution would result in re-running the applications, which will cost time, money and energy. In this paper we presented a proactive fault tolerance (FT) approach to HPC systems in the cloud to reduce the wall-clock execution time and dollar cost in the presence of faults. We also developed a generic FT algorithm for HPC systems in the cloud. Our algorithm does not rely on a spare node prior to prediction of a failure. We also developed a cost model for executing computation-intensive applications on HPC systems in the cloud. We analysed the dollar cost of provisioning spare nodes and checkpointing FT to assess the value of our approach. Our experimental results obtained from a real cloud execution environment show that the wall-clock execution time and cost of running computation-intensive applications in cloud can be reduced by as much as 30%. The frequency of checkpointing of computation-intensive applications can be reduced up to 50% with our FT approach for HPC in the cloud compared with current FT approaches.