Supercomputing Resources Research Articles

Brain Tumor Recognition Using Artificial Intelligence Neural-Networks (BRAIN): A Cost-Effective Clean-Energy Platform

Brain tumors necessitate swift detection and classification for optimal patient outcomes. Deep learning has been extensively utilized to recognize complex tumor patterns in magnetic resonance imaging (MRI) images, aiding in tumor diagnosis, treatment, and prognostication. However, model complexity and limited generalizability with unfamiliar data hinder appropriate clinical integration. The objective of this study is to develop a clean-energy cloud-based deep learning platform to classify brain tumors. Three datasets of a total of 2611 axial MRI images were used to train our multi-layer convolutional neural network (CNN). Our platform automatically optimized every transfer learning and data augmentation feature combination to provide the highest predictive accuracy for our classification task. Our proposed system identified and classified brain tumors successfully and efficiently with an overall precision value of 96.8% [95% CI; 93.8–97.6]. Using clean energy supercomputing resources and cloud platforms cut our workflow to 103 min, $0 in total cost, and a negligible carbon footprint (0.0014 kg eq CO2). By leveraging automated optimized learning, we developed a cost-effective deep learning (DL) platform that accurately classified brain tumors from axial MRI images of different levels. Although studies have identified machine learning tools to overcome these obstacles, only some are cost-effective, generalizable, and usable regardless of experience.

A Study on the Operational Efficiency of National Supercomputer Resources Based on DEA Model

As a measure to improve the operational efficiency of the Korean supercomputer joint utilization resources, an efficiency analysis using DEA was conducted to derive the improvement level of the input and output factors of the operation plans of seven specialized centers. The factors that constitute the objective function of efficiency were identified through the operation plans submitted by the specialized centers, and DEA analyzed the improvement factors and levels of inefficient specialized centers through the CCR and BCC models, and then conducted an additional super-efficiency analysis to compare the relative efficiency levels among the efficiency groups. As a result of the analysis, the inefficiency of specialized centers in the meteorological/climate/environmental fields was identified from the reference group, and the need for improvement and the level of improvement of input and output variables were identified. In addition, the authors suggested ways to improve the establishment of resources and operational targets, such as a joint utilization ratio to secure financial efficiency at the government level in establishing a joint utilization resource.

Исследование электровихревого течения между плоскостями с помощью различных вычислительных подходов

Electrovortex flows arise when an electric current of varying density passes through a well-conducting fluid (e.g., acid or metal melt). In such a case, the electric current generates a magnetic field, which leads to the Lorentz force causing swirling currents of the medium. There are different methods of theoretical study of such currents. As a rule, to avoid the necessity to find the pressure dependence on coordinates, the variables "vector potential of velocity - swirl" ("scalar current function – swirl" in the case of axisymmetric flows) are used. In such a case, it is quite effective to use automodel variables, which allow to reduce the dimensionality of the problem. In this case, the solution for the introduced function can be sought in the form of an expansion by the electro-vortex flow parameter proportional to the square of the magnetic Reynolds number. Also, this solution can be obtained numerically, for example, using finite-difference methods. Nowadays, more and more often the solutions are investigated by means of direct numerical modeling methods, when no automodel approximations are made, which reduce the accuracy of the solution. Nevertheless, in such a case the number of computations can be quite large and requires the use of supercomputer resources. A separate difficulty is presented by the boundary conditions: for example, for the velocity vector potential we obtain a fourth-order equation, which imposes significant restrictions on the time steps in the evolution equation. The problem can be avoided by using approximate boundary conditions, but this again reduces the accuracy of the solution. In this paper, using the example of electro-vortex flow between planes, the solutions that can be obtained using the various computational approaches mentioned above are examined. The results obtained are compared, and they are also compared with analytical approximations.

Physics-Based Watercraft Simulator in Virtual Reality

In this paper, watercraft and ship simulation is summarized, and the way that it can be extended through realistic physics is explored. A hydrodynamic, data-driven, immersive watercraft simulation experience is also introduced, using the Unreal Engine to visualize a Landing Craft Utility (LCU) operation and interaction with near-shore waves in virtual reality (VR). The VR application provides navigation scientists with a better understanding of how coastal waves impact landing operations and channel design. FUNWAVE data generated on the supercomputing resources at the U.S. Army Corps of Engineers (USACE) Engineering Research and Development Center (ERDC) are employed, and using these data, a graphical representation of the domain is created, including the vessel model and a customizable VR bridge to control the vessel within the virtual environment. Several dimension reduction methods are being devised to ensure that the FUNWAVE data can inform the model but keep the application running in real time at an acceptable frame rate for the VR headset. By importing millions of data points output from the FUNWAVE version 3.4 software into Unreal Engine, virtual vessels can be affected by physics-driven data.

A study on demand management plans for National Supercomputer resources

In response to the rapid surge in AI computing demand at a large scale, the requirement for effective management of national supercomputing resources has emerged in the Republic of Korea. The objective is to not just amplify supply in accordance with demand but also regulate demand based on supply availability. The primary rationale behind this pertains to our complete reliance on purchasing foreign computing resources. Given that all expenditures for resource development are drawn from the government budget, numerous economic constraints come into play. Therefore, this study outlines a plan for managing the demand of national supercomputing resources and assesses its impact across various scenarios. By utilizing operational data from the National Center's supercomputing operations, we evaluated the potential and degree of enhancement in operational efficiency using Data Envelopment Analysis. Additionally, we scrutinized the effects of enhancing operational efficiency for each scenario by applying the Valley-Filling demand-management approach. To quantitatively compare operational effects in each scenario, load factor metrics were employed. The analysis revealed both the tangible improvement in operational efficiency due to demand management and crucial factors pivotal for formulating future demand management strategies. The outcomes are expected to serve as a primary guide for shaping a demand management system and devising prospective action plans.

Roadmap on electronic structure codes in the exascale era

Abstract Electronic structure calculations have been instrumental in providing many important insights into a range of physical and chemical properties of various molecular and solid-state systems. Their importance to various fields, including materials science, chemical sciences, computational chemistry, and device physics, is underscored by the large fraction of available public supercomputing resources devoted to these calculations. As we enter the exascale era, exciting new opportunities to increase simulation numbers, sizes, and accuracies present themselves. In order to realize these promises, the community of electronic structure software developers will however first have to tackle a number of challenges pertaining to the efficient use of new architectures that will rely heavily on massive parallelism and hardware accelerators. This roadmap provides a broad overview of the state-of-the-art in electronic structure calculations and of the various new directions being pursued by the community. It covers 14 electronic structure codes, presenting their current status, their development priorities over the next five years, and their plans towards tackling the challenges and leveraging the opportunities presented by the advent of exascale computing.

Natural products subsets: Generation and characterization

Natural products are attractive for drug discovery applications because of their distinctive chemical structures, such as an overall large fraction of sp3 carbon atoms, chiral centers (both features associated with structural complexity), large chemical scaffolds, and diversity of functional groups. Furthermore, natural products are used in de novo design and have inspired the development of pseudo-natural products using generative models. Public databases such as the Collection of Open NatUral ProdUcTs and the Universal Natural Product database (UNPD) are rich sources of structures to be used in generative models and other applications. In this work, we report the selection and characterization of the most diverse compounds of natural products from the UNPD using the MaxMin algorithm. The subsets generated with 14,994, 7,497, and 4,998 compounds are publicly available at https://github.com/DIFACQUIM/Natural-products-subsets-generation. We anticipate that the subsets will be particularly useful in building generative models based on natural products by research groups, particularly those with limited access to extensive supercomputer resources.

Bringing Quantum Mechanics to Coarse-Grained Soft Materials Modeling

Fundamental knowledge gaps are endemic in our understanding of how emergent properties of soft materials are linked to the quantum mechanical (QM) world. The limitations of current QM modeling paradigms inhibit the understanding and design of classes of soft materials for which QM phenomenology is critical. At its root, these limitations derive from the seemingly innocuous premise of requiring all atomic positions to solve the molecular Schrödinger equation, which necessitates supercomputing resources to incorporate even simple QM phenomenology into small (∼nm) systems of soft materials. Here, we review emerging efforts to overcome these challenges through the development of electronic prediction models that operate at the coarse-grained resolution. We motivate the origins of this new computational paradigm, denoted electronic coarse-graining (ECG), discuss its relationship to existing molecular modeling frameworks, and describe recent successes of ECG and related models for soft materials. Importantly, we highlight the classes of soft materials where ECG models can be potentially transformative.

A Study on the Designation Institution for Supercomputer Specialized Centers in Republic of Korea

In Korea, specialized centers are designated for 10 strategic fields for the purpose of jointly utilizing supercomputer resources at the national level. Based on the “National Supercomputing Innovation Strategy,” it plans to select 10 centers in three stages by 2030, and has now completed the designation of the first-stage specialized centers in 2022. With the second designation in 2024 ahead, it is urgent to review and improve the existing designation institution for fairer and more effective selection of specialized centers. Therefore, this paper analyzed the influence of evaluation items and the influence of evaluation items on evaluation results by using logistic regression analysis and network centrality analysis to prepare improvement plans for the existing evaluation model. As a result of the analysis, improvement measures were derived, such as subdividing evaluation items with low impact, expanding the items, and lowering the allotment of evaluation items with low impact.

A Study on Institution Improvement Plans for the National Supercomputing Joint Utilization System in South Korea

The purpose of this paper is to discover institutional gaps in the supercomputing joint utilization system that the government is actively promoting as an alternative to the problem of shortage of domestic supercomputing resources. The institutional gaps were discovered by examining the current status of laws, top-level plans, and operating guidelines related to the current joint utilization system and matching them with problems or issues that need to be resolved socially. The improvement plan for the institutional gaps was derived at a level that can be reflected in the operating guidelines of the Specialized Center and Unit Center so that the performing subject constituting the joint utilization system can directly participate and solve it. In the future, for the effective operation of the joint utilization system, we plan to promote the domestic market through the diffusion of research results and secure external technological competitiveness by reflecting the contents of institution improvement.

Comparison of Three Job Mapping Algorithms for Supercomputer Resource Managers

Performance of supercomputer depends on the quality of resource manager, one of its functions is assignment of jobs to the nodes of clusters or MPP computers. Parts of parallel programs interact with each other with different intensity, and mapping of program to supercomputer nodes influence efficiency of the run. At each program run graph representing application program is to be mapped onto graph of nodes representing a subset of computer system. The both graphs are not known beforehand, hence the mapping must be done in reasonable time while scheduling resources. Three mapping algorithms were explored: parallel versions of simulated annealing, genetic and composite algorithms. A set of experimental runs with different algorithms parameters was performed, comparison of mapping quality and runtime was made, and suggestions on applicability of algorithms for resource managers were provided.

MemXCT: Design, Optimization, Scaling, and Reproducibility of X-Ray Tomography Imaging

This work extends our previous research entitled &#x201C;MemXCT: Memory-centric X-ray CT Reconstruction with Massive Parallelization&#x201D; that was originally published at SC19 conference (Hidayeto&#x011F;lu <i>et al.</i>, 2019) with reproducibility of the computational imaging performance. X-ray computed tomography (XCT) is regularly used at synchrotron light sources to study the internal morphology of materials at high resolution. However, experimental constraints, such as radiation sensitivity, can result in noisy or undersampled measurements. Further, depending on the resolution, sample size and data acquisition rates, the resulting noisy dataset can be in the order of terabytes. Advanced iterative reconstruction techniques can produce high-quality images from noisy measurements, but their computational requirements have made their use an exception rather than the rule. We propose a novel memory-centric approach that avoids redundant computations at the expense of additional memory complexity. We develop a memory-centric iterative reconstruction system, MemXCT, that uses an optimized SpMV implementation with two-level pseudo-Hilbert ordering and multi-stage input buffering. We evaluate MemXCT on various supercomputer architectures involving KNL and GPU. MemXCT can reconstruct a large (11K&#x00D7;11K) mouse brain tomogram in 10 seconds using 4096 KNL nodes (256K cores). The results presented in our original article at the SC19 were based on large-scale supercomputing resources. The MemXCT application was selected for the Student Cluster Competition (SCC) Reproducibility Challenge and evaluated on a variety of cloud computing resources by universities around the world in the SC20 conference. We summarize the results of the top-ranked SCC Reproducibility Challenge teams and identify the most pertinent measures for ensuring the reproducibility of our experiments in this article.

How to Assess the Quality of Supercomputer Resource Usage

How to Assess the Quality of Supercomputer Resource Usage

Supercomputer Search for the New Inhibitors of the Coagulation Factor XIIa

A virtual structure-oriented screening of candidates for inhibitors of blood coagulation factor XIIa was carried out. This coagulation factor is one of the most promising therapeutic targets for the development of anticoagulants without disturbing normal hemostasis. The screening was carried out in the database of organic compounds of the Voronezh State University, consisting of more than 19 thousand molecules. At the first stage of virtual screening, ligands were positioned in the active center of factor XIIa using the SOL docking program. At the second stage, for the best ligands, the protein-ligand binding enthalpy was calculated using the MOPAC program and the PM7 quantum-chemical method, taking into account the solvent in the COSMO continuum model. All calculations were carried out using the supercomputing resources at Lomonosov Moscow State University. In total, more than 30 thousand ligand conformers are docked, and for more than 400 of the best of them, the protein-ligand binding enthalpy was calculated. 16 compounds are selected as the most promising candidates for subsequent in vitro testing of their ability to inhibit factor XIIa.

Supercomputer model of dynamical dusty gas with intense momentum transfer between phases based on OpenFPM library

The paper considers the solution of model gas-dynamic problems (propagation of plane sound wave, one-dimensional shock tube problem, three-dimensional problem of a point explosion in a continuous medium) in the case of a gas-dust medium. The interaction of dust and gas was taken into account using the IDIC method within the SPH method used to solve gas-dynamic equations. An important feature of the work is the use of the open computational package OpenFPM, which makes it easy to carry out parallel computations. The main advantage of this package is the ready-made (implemented by the authors of the package) and intuitive, automatically parallelizable vector data structures, the use of which is identical both in the case of calculations on a personal computer and in the case of using supercomputer resources. The paper analyzes the efficiency of parallelization of numerical solutions of the considered problems.

The Prospects of Development in Russia of High-Performance Computing and Predictive Modeling in Modern Technologies.

This article is dedicated to the development and application of supercomputers and supercomputer technologies, which constitute a major factor in scientific, technical, and social advance, for solving problems that require large computing capacities. Trends in the use of supercomputers are considered, and the level of equipping advanced countries with supercomputer resources is discussed. Examples of tasks the solution of which is impossible without using computing capacities of tens of petaFLOPS are given. This article was prepared on the basis of a report presented at the meeting of the Presidium of the Russian Academy of Sciences on February 16, 2021.

JUSUF: Modular Tier-2 Supercomputing and Cloud Infrastructure at Jülich Supercomputing Centre

JUSUF is a petaflop supercomputer operated by Jülich Supercomputing Centre at Forschungszentrum Jülich as a European supercomputing and cloud resource. JUSUF was funded via the ICEI project and especially serves the Human Brain Project and PRACE via ICEI and the Fenix Research Infrastructure. The system consists of two parts, an HPC cluster partition and an Infrastructure-as-a-Service cloud partition. The system entered production phase in spring 2020. It is based on the Bull X400 product family with AMD Rome processors, partially accelerated by Nvidia V100 GPUs, and Nvidia Mellanox HDR InfiniBand.

JUWELS Cluster and Booster: Exascale Pathfinder with Modular Supercomputing Architecture at Juelich Supercomputing Centre

JUWELS is a multi-petaflop modular supercomputer operated by Juelich Supercomputing Centre at Forschungszentrum Juelich as a European and national supercomputing resource for the Gauss Centre for Supercomputing. In addition, JUWELS serves the Earth system modeling community and the AI community within the Helmholtz Association as well. JUWELS currently consists of two modules. The first module deployed in 2018 is the so-called Cluster module. The Cluster is a BullSequana X1000 system with Intel Xeon Skylake-SP processors and Mellanox EDR InfiniBand. The second module deployed in 2020 is the so-called Booster module. The Booster is a BullSequana XH2000 system with 2nd generation AMD EPYC processors, NVIDIA Ampere GPUs and NVIDIA/Mellanox HDR Infiniband. This paper describes in detail the architecture of the system from a users perspective, and additionally provides further insights into the administrative infrastructure used to operate the supercomputer.

Quantifying the benefits of in-time and in-place responses to remediate acute LNAPL release incidents

Acute large volume spills from storage tanks of petroleum hydrocarbons as light non aqueous phase liquids (LNAPLs) can contaminate soil and groundwater and may have the potential to pose explosive and other risks. In consideration of an acute LNAPL release scenario, we explore the value of a rapid remediation response, and the value of installing remediation infrastructure in close proximity to the spill location, in effecting greater recovery of LNAPL mass from the subsurface. For the first time, a verified three-dimensional multi-phase numerical framework and supercomputing resources was applied to explore the significance of in-time and in-place remediation actions. A sand aquifer, two release volumes and a low viscosity LNAPL were considered in key scenarios. The time of commencement of LNAPL remediation activities and the location of recovery wells were assessed requiring asymmetric computational considerations. The volume of LNAPL released considerably affected the depth of LNAPL penetration below the groundwater table, the radius of the plume over time and the recoverable LNAPL mass. The remediation efficiency was almost linearly correlated with the commencement time, but was a non-linear function of the distance of an extraction well from the spill release point. The ratio of the recovered LNAPL in a well located at the centre of the spill/release compared to a well located 5 m away was more than 3.5, for recovery starting only 7 days after the release. Early commencement of remediation with a recovery well located at the centre of the plume was estimated to recover 190 times more LNAPL mass than a one-month delayed commencement through a well 15 m away from the centre of the LNAPL plume. Optimally, nearly 40% of the initially released LNAPL could be recovered within two months of commencing LNAPL recovery actions.

Students in a Course-Based Undergraduate Research Experience Course Discovered Dramatic Changes in the Bacterial Community Composition Between Summer and Winter Lake Samples

Course-based undergraduate research experience (CURE) courses incorporate high-impact pedagogies that have been shown to increase undergraduate retention among underrepresented minorities and women. As part of the Building Infrastructure Leading to Diversity program at the University of Detroit Mercy, a CURE metagenomics course was established in the winter of 2019. Students investigated the bacterial community composition in a eutrophic cove in Lake Saint Clair (Harrison Township, MI, United States) from water samples taken in the summer and winter. The students created 16S rRNA libraries that were sequenced using next-generation sequencing technology. They used a public web-based supercomputing resource to process their raw sequencing data and web-based tools to perform advanced statistical analysis. The students discovered that the most common operational taxonomic unit, representing 31% of the prokaryotic sequences in both summer and winter samples, corresponded to an organism that belongs to a previously unidentified phylum. This result showed the students the power of metagenomics because the approach was able to detect unclassified organisms. Principal Coordinates Analysis of Bray–Curtis dissimilarity index data showed that the winter community was distinct from the summer community [Analysis of Similarities (ANOSIM) r = 0.59829, n = 18, and p < 0.001]. Dendrograms based on hierarchically clustered Pearson correlation coefficients of phyla were divided into a winter clade and a summer clade. The conclusion is that the winter bacterial population was fundamentally different from the summer population, even though the samples were taken from the same locations in a protected cove. Because of the small class sizes, qualitative as well as statistical methods were used to evaluate the course’s impact on student attitudes. Results from the Laboratory Course Assessment Survey showed that most of the respondents felt they were contributing to scientific knowledge and the course fostered student collaboration. The majority of respondents agreed or strongly agreed that the course incorporated iteration aspects of scientific investigations, such as repeating procedures to fix problems. In summary, the metagenomics CURE course was able to add to scientific knowledge and allowed students to participate in authentic research.