초고성능컴퓨팅 계산과학공학 연구 및 선도기술 개발

Abstract

□ 사업의 목표: 세계 수준의 슈퍼컴퓨팅 R&D․시범서비스 도출 및 독자 연구개발 역량 확보

□ 주요사업 내용 ○ 초고성능컴퓨팅 활용 계산과학공학 선도 연구 - R&D 효율화를 위한 계산과학연구 선도 - 세계 수준의 연구 성과 도출(기술혁신, 과학적 발견) - 거대 컴퓨팅 및 데이터 활용의 국제 공동연구의 R&D 파트너 구축 - 계산과학 연구 분야 신규 창출 인큐베이터 수행

○ 거대 컴퓨팅 가시화 기술 개발 - 세계 수준의 거대 컴퓨팅 계산과학공학 가시화 기술 개발로 R&D 효율화 및 선도 기술 확보 - 기술의 실용화 및 기술이전으로 거대 컴퓨팅 가시화 기술 활용확대

○ 계산과학공학 선도 기술 확산을 위한 국내외 공동연구체계 구축 및 육성 - 선도 기술 확산을 위한 공동연구체계 구축 및 육성 - 차세대 계산과학공학 플랫폼 개발을 위한 선도연구 주제 탐색

( 출처 : 초록 4p )

Ⅳ. Result of the study ㅇ Studies on computational science using supercomputer - The achievement of the world-class results on bio-physics beyond theory and experiments => 3 top-20% SCI papers published - The study of particle physics using data-intensive supercomputer => 1 top-20% SCI and 1 SCI papers published - To make new area on condensed matter physics => 2 papers published - To achieve research fund to collaborative with Korea Astronomy and Space Science Institute(KASI)

ㅇ The massive data visualization technology development - The performance improvement for developing world class massive data visualization technology • The visualization performance achievement for 2TB data is 1.3 times faster than VisIt, which has the world’s best large data visualization record. • The visualization performance for complex data has been improved 105 times compared to last year’s performance. • The unstructured grid volume rendering performance has been improved 300% compared to VTK, which is an open-source visualization development toolkit

- The usability improvement for expanding massive data visualization technology utilization • The user-execution time has been saved 47% by developing the VR user interface based on smart devices. • We’ve developed communication protocols between the VR system and the smart device. • We’ve implemented the widgets for developing user interfaces. • The development of real-time high caliber rendering technology has improved the realistic feelings of simulation data visualization results. • The SUS of the user interface for 2nd value generation functions has achieved 1.43 times higher than EnSight, the commercial post-processing tool. • The SUS of graph functions has achieved grade A(excellent).

- The Joint research and technology transfers for improving technology availability • The process time of the multi-dimensional index data searching algorithm based on GPU has improved 100 times. (Joint research) • The memory usage has saved maximum 90% by developing SAMR data specialized data structure for GPU volume rendering. • 2 cases of paid-in technology transfer using research results has been achieved.

ㅇ Establishment and promotion of domestic and international collaborative research community for spreading advanced technology of computational science and engineering - Establishment and promotion of collaborative research community for spreading advanced technology of computational science and engineering • Transferring core technology of the EDISON platform including its core parts by the ODA program between KISTI(Korea) and ICST (Vietnam) • Acquiring collaborative opportunities for developing world-class supercomputing core technology and advanced computational science platform • Carrying out a variety of joint programs such as PRAGMA, researcher exchange, and workshops, for expanding global user community and mining collaborative research under the same goal of constructing a simulation-based cyber-learning system between Korea and Taiwan • Establishing a joint lab between the Center of Supercomputing Convergence R&D at KISTI and the Supercomputing Center at UNIST for inter-collaboration to develop platform technology together and exchange supercomputing-related research achievements

- Exploring advanced research topics for developing the next-generation computational science and engineering platform • Development of data-driven simulation platform • Development of a dynamic resource-allocation technique employing user logs • Development of a dynamic job-scheduling technique • Development of a software-based file-deduplication technique driven by repetitive simulations • Development of an efficient storage management technique adaptive to simulation execution by numerous users

( 출처 : SUMMARY 13p )