[보고서]슈퍼컴퓨팅 활용 확대를 위한 과학·공학 분야 거대문제 해결·지원

염민선

보고서 정보

주관연구기관

한국과학기술정보연구원
Korea Institute of Science and Technology Information

연구책임자

염민선

참여연구자

강궁원 , 강지훈 , 강효진 , 권오경 , 김영훈 , 김인 , 김한슬 , 류훈 , 문명환 , 문지협 , 박기홍 , 백효정 , 서상재 , 안설아 , 안아름 , 여인성 , 유석종 , 육진희 , 윤태호 , 이재학 , 이준학 , 이홍석 , 정민중 , 정영균 , 정찬석 , 정희진 , 조기현 , 조용성 , 조지호 , 최성환 , 현영환 , 박찬 , 서진석 , 허태상 , 황순욱 , 최장원 , 정용환 , 고명주 , 최윤근 , 이식 , 허영주 , 류범종 , 김병규 , 심형섭 , 손지성 , 이상환 , 송사광 , 김성찬 , 안부영 , 한성근 , BUI KHAC HOAI NAM , 한기훈 , 차원철 , 최정일 , 오상윤

보고서유형

2단계보고서

발행국가

대한민국

언어

한국어

발행년월

2019-12

과제시작연도

2019

주관부처

과학기술정보통신부
Ministry of Science and ICT

등록번호

TRKO202000006303

과제고유번호

1711097340

사업명

한국과학기술정보연구원연구운영비지원(R&D)(주요사업비)

DB 구축일자

2020-07-29

키워드

계산과학공학.슈퍼컴퓨팅 기술.성능 최적화.거대문제.병렬화.Computational Science and Engineering.Supercomputing Technology.Performance Optimization.Grand-challenge Problem.Parallelization.

표 Schematic of cache-aware shuffling optimization
표 List of 2019 calls for collaboration of performance optimization
표 Supporting program list of 2019 parallel performance optimization
표 Multi-level parallelization for LANZCOS algorithm
표 Wave analysis problem
표 High-parallel algorithm for thermal-fluid analysis
표 Test of Poisson solver
표 Visualization of Acoustic Sensor Data Using (Visual-Check: C-2019-040405)
표 Visualization of Actual Sensor Data using KISTI Pattern Analysis
표 (a) World's largest cosmological numerical experiment (Box size > 100 Mpc/h)
표 Detail of normal cells and cancer cells
표 Large-scale simulation for electric device design
표 Flow of CADD(Computer-Adided Drug Design) and virtual screening
표 InAs based Nanowire sublimation
표 Fullerence(C60, C70)
표 DNA analysis sensor
표 Group cooperation strategy for brain disease
표 FEA of Seismic propagation
표 Example of large-scale thermal and/or fluid simulation
표 Nuclear chart
표 Low energy physics profiling [KISTI] (leftplot plot), High Energy Physics Profiling [Fermilab] (right plot)
표 Bacterial chromosome and molecular crowding/clustering
표 Mathematical formula and schematic diagram of t-J model Hamiltonian
표 Numerically predicted phase diagram of high-temperature superconducting cuprates
표 Research Strategies and Methods
표 Relationship between sub-projects (R&D programs)
표 Annual status of performance optimization and parallelization
표 Best papers by parallel optimization technical support
표 parallel performance test of turbulence simulation by using maximum 6,144cores
표 Research of OLCF
표 RI-MP2 of (C150H30)2 calculation by K-computer of 80,199 nodes
표 Gordon Bell award list in peak performance
표 Application result of communication method by layer using ResNet-50 model (a) Image throughput per second by 2048 KNL node/ (b) Image throughput per second by 4096 KNL node
표 Image throughput per second of communication method by layer(top 10%)
표 Cache-aware Shuffling 기법
표 Algorithm of Parallel and Scalable Library for TDMA(PaScaL_TDMA)
표 Schematic of a tridiagonal system. (a) A partitioned sub-matrices, (b) the modified submatrices, and (c) a reduced tridiagonal system
표 Schematic of (a) the present method using MPI Alltoall for many tridiagonal systems, and (b) a communication method using MPI Gather for a single tridigonal system
표 The strong scalability of the conventional method and the proposedmethods of PaScaL_TDMA single and PaScaL_TDMA many
표 The scalability of PaScaL_TDMA many for solving the three-dimensional heat equation. (a) Strong scalability and (b) weak scalability
표 The strong scalability for VDMPM-STD with PaScaL_TDMA (a) Comparison with previous solver (b) Strong scalability for omputational partsof VDMPM-STD
표 Thread-specific scalability of RAMSES code in Nurion 2,500 node
표 (a) SIO performance comparison with PFS and BB (512x512x512 grid, 17GB file generation) (b) WRF performance comparison with PFS and BB(input file size: 4.8GB, output file size: 4.7GB, 128 node usage)
표 Total elapsed time and ratio of diagonalization on a) KNL ｂ) SKL nodes
표 ResNet-50 image throughput using per-layer communication optimization using (a) 2048 and (b) 4096 KNL nodes
표 Image throughput improvement using per-layer communication optimization
표 Strong scalability of RAMSES codes at 2,500 nodes of Nurion KNL
표 (a) Performance comparison of SIO between Parallel File System(PFS) and Burst Buffer(BB) (512x512x512, 17GB File) (b) Performance comparison of WRF between PFS and BB (Input file: 4.8GB, Output file: 4.7GB, 128 nodes)
표 (a) Schematic of 2D domain decomposition (b) 3D Domain decomposition using Parallel TDMA solver
표 (a) Strong scalability of LES code at KISTI Nurion supercomputer (1537x641x4097 number of grids) (b) Performance comparison of 2D and 3D domain decomposition (1537x641x4097, blue: 2D, red: 3D) (c) CGNS based parallel visualization
표 Method of parallel input file generation for FDS parallel execution
표 Flow chart
표 Parallel benchmark test results for FDS
표 Performance comparison of NUFESOL program (Blue: Nurion KNL, Red: Nurion CPU-only, Green: baseline)
표 Analytical pipeline data flow using distributed processing
표 Velocity comparison of analysis pipeline depending on file size
표 Performance comparison depending on SDedup Resource quota
표 CLI application mode
표 Nurion parallelization test of windAroundBuildings example
표 OpenFOAM education program operation(Data school cooperation)
표 Presentation of KISTI OpenFOAM service application at OpenFOAM Workshop(Germany)
표 Summary of supercomputer resource share
표 1st / 2nd list of support tasks of Innovation Support Program(giant research)
표 1st / 2nd list of support tasks of Innovation Support Program(innovative growth)
표 Experimentally demonstrated optoelectronic characteristics of Metal-Halide-Perovskite (MHP) materials: (a) Emission wavelength of various MHP composition (b) Red-shift stemming from light-induced phase separation
표 (Left) Structures of simulated supercells (Quantum Dots / Nanowires / Nanoplatelets): ‘L’ symbol indicates the confinement size of supercells, which ranges from 3nm to 20nm (Right) Intel official case study report that reveals the strong scalability of large-scale tight-binding electronic structure calculations up to 7.5PF (2,500 KNL nodes)
표 (a) Cover of the research article published in Journal of Physical Chemistry Letters (IF 7.8) (b) Simulation results (1) indicate that it is possible to reduce the red-shift by a factor of 4 with no remarkable change in band gap energy and (2) show strong connections to reported experimental results
표 CNN based Inverse design of a nanohole metasurface
표 Positive Q-invariant of heat flux gradient tensor with temperature contour(Pr =0.021,Ra=2X109)
표 Temperature contour (a) Pr = 0.7, Ra=5×108 (b) Pr = 0.7, Ra=2×1010 (c) Pr = 0.021, Ra=107 (d) Pr = 0.021, Ra=5×108
표 The barycentric map of wall-normal profile extracted from maximum JPDF location (a) Pr = 0.7, Ra=5×108 (b) Pr = 0.7, Ra=2×1010 (c) Pr = 0.021, Ra=107 (d) Pr = 0.021, Ra=5×108
표 Current and target simulation conditions
표 Result of HR5 describing the formation of primitive galaxies, clusters of galaxies and massive structures
표 Result of HINT LCDM at z=2(10 billion years ago)
표 Nuclear reactions at low energies
표 Effect of three body forces
표 ValidateFastq result screen for unequal length of files for Paired-end fastq files
표 ValidateFastq result screen for qualified paired-end fastq file pair
표 Snakemake pipeline for VerifyBamID
표 QC result screen for checking cross-contamination by using VerifyBamID
표 Pathoscope pipeline
표 Moustafa et al.‘s blast-base pipeline
표 Parallel efficiency for a unit genomic analysis
표 Fastqc result graph for Batch 4 data
표 Fastqc result graph for Batch 5 data
표 Depth of coverage graph for Batch 4 data
표 Depth of coverage graph for Batch 5 data
표 Insert size graph for Batch 4 data
표 Insert size graph for Batch 5 data
표 List of samples with incorrect gender description detected by QC pipeline
표 List of families with incorrect familial information detected by QC pipeline
표 Statistics of de novo variant identified
표 Result of the phased de novo variant
표 Information of the identified de novo CNVs
표 IGV plot for 7q34 deletion of 1323-3 sample
표 Circos plot for de novo mutations from the whole ASD probands
표 Pie chart for the identified de novo variant’s type
표 DenovolyzeR analysis result table
표 Correlation between the number of de novo SNV and parents’ age at child’s birth
표 Correlation between number of genes affected by de novo variants and IQ
표 Mutation signature analysis results for de novo SNVs from ASD samples and sibling samples
표 Disease impact score analysis result
표 Schematic diagram for Alzheimer’s disease research by using low-frequency somatic variant calling pipeline
표 Statistics of the brain sample doners
표 Protocol of DNA isolation for the brain and the blood samples
표 Schematic diagram of low-frequency somatic variant calling pipeline
표 The contingency table of targeted sequencing results and low-frequency variant calling pipeline results
표 Correlation between allele frequencies of WEseq and TAseq
표 Comparison of No. of the low-frequency somatic mutations in brain and blood samples from Alzheimer patients and normal person
표 Comparison of mean VAF of somatic SNVs in brain and blood samples from Alzheimer patients and normal person
표 Comparison of the ratio of somatic SNVs’ type in brain and blood samples from Alzheimer patients and normal person
표 Number of somatic SNVs in brain samples according to the age of the sample
표 Number of somatic SNVs in blood samples according to the age of the sample
표 Mutational signature analysis results for somatic variants identified from brain samples
표 Mutational signature analysis results for somatic variants identified from blood samples
표 Pathway enrichment test results for the tau aggregation related pathways
표 The presence of pathogenic germline risk factors and putatively pathogenic somatic mutations across all AD and non-AD individuals
표 Categorization of pathogenic germline and somatic mutation carriers among AD patients
표 Pathogenic scoring and text- mining results of brain somatic mutations found in AD
표 Effect of p.Thr152Met mutation on PPIC domain of PIN1 computed by using SDM2
표 Effect of Pin1 knockdown on the oligomerization of tau
표 Media broadcase example of Alzheimer’s disease research in KISTI
표 10 million disease trajectory tracking visualizations: each line represents the patient's re-entry follow-up path and the middle node represents the diagnosis
표 Complication pattern presented with big data (Daejeon MBC, etc.)
표 in vitro functional validation of site-directed mutation
표 Classification of secretory and non-secretory HPs
표 Distribution of number of aligned sequences of multiple sequence alignments of secretory HPs (blue) and non-secretory HPs (red). Three shared motifs in secretory HPs are identified (arrow). Homeodomain regions are boxed
표 Comparison of sequence logos of homeodomain region for our MSA to those for Pfam
표 Amino acid sequences in the positions pointed by multiple sequence alignment
표 Brain single cell clustering assay for mental illness mouse models
표 Association of SHANK3 gene mutations with various neuropsychiatric diseases
표 Single cell isolation from mouse striatum and medial prefrontal cortical brain tissue
표 QC Results for Single Cell RNA Libraries
표 scRNA-seq analysis workflow
표 scRNA-seq Assay Protocol Specs Comparison
표 scRNA-seq Assay Protocol Accuracy Comparison
표 10x Genomics Single Cell Sequencing Assay Platform
표 Cell Ranger Analysis Pipeline
표 Mapping analysis result through Cell Ranger
표 Result of applying data QC filtering
표 Single Cell Clustering Assay
표 Single-cell cell-type clustering result
표 Expression pattern analysis of marker genes by cell type
표 Seurat-provided single cell integration algorithm
표 Integrated Analysis of Structural Normal and Mutant Cells
표 OPC cell-specifically expressed C1ql1 gene
표 Scoring function for calculating gene-gene similarity and its distribution for all gene pairs
표 Comparison of features from different network methods
표 Accuracy comparison of different networks
표 Diagram of collaborative research between KISTI-Georgia Tech.-BOSCH
표 Schematic diagram of DNA sequencing sensor
표 Interface structure-dependent change of the insulating properties of metal-insulator-metal junction
표 Hydrophilic/hydrophobic properties of crystal violet molecules and their behavior in solvents
표 (a) Suggested docking positions of a Catechin ligand to the Cathepsin C protein and its (b) root-mean-squared distances during dynamics simulations and (c) heatmap representation for active docking sites
표 Sublimation simulations on InAs nanowire surfaces using molecular dynamics simulations
표 Surface energies of InAs nanowires on various facets
표 Partial views of the calculated proton NMR spectra of (a) aromatic and (b) aliphatic regions of 1 and 1:C60. (c) Calculated vertical transitions, (d) energy level diagrams and molecular orbitals
표 Optimized structures and relative energies (in kJ/mol) of (a) 1 and (b) 1:C60 calculated at the B3LYP with a 6-31G(d) and B3LYP with a 6-31G(d)/PCM(toluene)
표 Pictorial representation of construction of quantum chemical database through high throughput computing
표 Press release for material development using supercomputing
표 (a) Nanowire supercells described with periodic boundary conditions (b) Atomic structure in a Wurtzite & Zincblende phase. For each crystal phase, 4 supercells in are simulated to investigate the effects dopant randomness on nanowire electronic structures
표 (a) Sensitivity of Piezoresistivity to external uniaxial tensor and (b) fluctuation of corresponding resistivity observed in both crystal phases (c) Simulation results present strong evidences that doping process increases the sensitivity of piezoresistivity, which can be figured out from the Density of State (DOS) distributions in nanowire supercells of both phases
표 Bacterial chromosome and molecular crowding/clustering
표 The purpose of high energy physics
표 Group convergence study by experiment-theory-simulation in the field of high energy physics
표 Simulation Tool Kit utilization for the field of high energy physics
표 Prediction of rare isotopes
표 The flow chart of signal and background in the 3 jet system
표 The CPU comparison of KISTI-4 and KISTI-5 supercomputers
표 The predicted U(1) phase diagrams in the plane of temperature and hole doping concentration for three different values of Heisenberg coupling constant, i.e., J/t = 0.2, 0.3, 0.4 on the 2 different unit cell lattice sizes 10×10 and 20×20
표 The predicted U(1) Phase diagram computed from 2-Dimensional t-J model Hamiltonian for J/t = 0.3 with the variation of unit cell lattice size (10×10, 20× 20, 30×30, 40×40, and 50×50)
표 The predicted SU(2) phase diagram computed from 2-dimensional t-J model Hamiltonian for J/t = 0.3 with the variation of unit cell lattice size (20×20, 30×30, 40×40, and 50×50)
표 Orbits for different spin configurations. Parallel to the orbital plane (Red), Anti-aligned (Blue). Gravitational waveforms emitted from black hole encounters (Bottom)
표 Top: Impact parameter vs Initial velocity or energy. Bottom: Comparisons with the parabolic approximation
표 (Left) Evolutions of the normalized Hubble parameters both in the Jordan frame (solid line) and in the Einstein frame (dashed line) (Right) Evolutions of effective equations of states of dark energy for both frames
표 the greybody factor in terms of the ratio between the spheroidal joining factors
표 Greybody factor curves of l=m=0 modes are compared for various rotation parameters a∗: Solid lines denote the graphs of the exact solutions and dashed lines those of approximate solutions in the low frequency limit
표 Six polarization modes: (a) breathing mode (b) longitudinal mode (c) vector-x mode (d) vector-y mode (e) plus mode (f) cross mode. The red, green, and blue colors indicate scalar, vector, and tensor modes, respectively. The circled dot in (b), (e), and (f) indicates respectively the wave propagating out of the page, and the right arrows in (a), (c), and (d) mean respectively the direction of wave propagation in the z-direction
표 The formalism of the exact six polarization modes of gravitational waves is constructed in terms of both the small metric perturbations and the Newman-Penrose scalars. The obtained formulae are applicable to any metric-compatible gravity theories whose gravitational waves propagate along either the null or non-null geodesics
표 Acoustic signal containing noise(top) vs. noise canceled signal(bottom)
표 Abnormal signal(right) compared to normal signal (left) measured from vehicle steering system
표 KISTI machine learning pattern recognition evaluation results

과제명(ProjectTitle) :	-
연구책임자(Manager) :	-
과제기간(DetailSeriesProject) :	-
총연구비 (DetailSeriesProject) :	-
키워드(keyword) :	-
과제수행기간(LeadAgency) :	-
연구목표(Goal) :	-
연구내용(Abstract) :	-
기대효과(Effect) :	-

내보내기 구분	파일저장 인쇄 메일전송
구성항목	기본정보 상세정보 관리번호, 제목(한글), 저자명(한글), 발행일자, 전자원문, 초록(한글), 초록(영문) 관리번호, 제목(한글), 제목(영문), 저자명(한글), 저자명(영문), 주관연구기관(한글), 주관연구기관(영문), 발행일자, 총페이지수, 주관부처명, 과제시작일, 보고서번호, 과제종료일, 주제분류, 키워드(한글), 전자원문, 키워드(영문), 입수제어번호, 초록(한글), 초록(영문), 목차
저장형식	Text(ASCII format) Excel format
메일정보	받는사람 (필수) @ 보내는사람 (선택) @ 제목 내용 KISTI 검색결과 이메일 서비스
안내	총 건의 자료가 검색되었습니다. 다운받으실 자료의 인덱스를 입력하세요. (1-10,000) 검색결과의 순서대로 최대 10,000건 까지 다운로드가 가능합니다. 데이타가 많을 경우 속도가 느려질 수 있습니다.(최대 2~3분 소요) 다운로드 파일은 UTF-8 형태로 저장됩니다. 파일의 내용이 제대로 보이지 않을실 때는 웹브라우저 상단의 보기 -> 인코딩 -> 자동선택 여부를 확인하십시오. ~ Text(ASCII format) Excel format

연합인증

슈퍼컴퓨팅 활용 확대를 위한 과학·공학 분야 거대문제 해결·지원
Solving grand-challenge problems in science and engineering to expand utilization of supercomputing 원문보기

초록 ▼

Abstract ▼

목차 Contents

표/그림 (182)

표/그림 (182)

연구자의 다른 보고서 :

참고문헌 (25)

연구과제 타임라인

관련 콘텐츠

원문 보기

이 보고서와 함께 이용한 콘텐츠

AI-Helper ※ AI-Helper는 오픈소스 모델을 사용합니다.

선택된 텍스트

연합인증

슈퍼컴퓨팅 활용 확대를 위한 과학·공학 분야 거대문제 해결·지원 Solving grand-challenge problems in science and engineering to expand utilization of supercomputing 원문보기

초록 ▼

Abstract ▼

목차 Contents

표/그림 (182) 모든 표/그림 보기

표/그림 (182) 슬라이드로 보기

연구자의 다른 보고서 :

참고문헌 (25)

연구과제 타임라인

전체(0) 논문(0) 특허(0) 보고서(0)

전체(0) 논문(0) 특허(0) 보고서(0)

관련 콘텐츠

원문 보기

이 보고서와 함께 이용한 콘텐츠

AI-Helper ※ AI-Helper는 오픈소스 모델을 사용합니다.

선택된 텍스트

슈퍼컴퓨팅 활용 확대를 위한 과학·공학 분야 거대문제 해결·지원
Solving grand-challenge problems in science and engineering to expand utilization of supercomputing 원문보기

표/그림 (182)

표/그림 (182)