HOKUSAI

the supercomputer bigwaterfall (bw) and the data science platform sailingship (ss)

RIKEN is Japan’s largest comprehensive research institution renowned for high-quality research in a diverse range of scientific disciplines.
The purpose of HOKUSAI is to contribute to scientific and technological research, and data processing and publishing by RKEN researchers.
- Used by more than 400 RIKEN researchers
- Fields: Particle and nuclear physics, Condensed matter physics, Physical Chemistry, Biophysics, Biophysics, Bioinformatics, and Others

The timeline of shared use computers in RIKEN Head Office for Information Systems and Cybersecurity

HOKUSAI BigWaterfall (BW)

ー BW-MPC: 840 nodes, 2.58 PFlops

CPU: Xeon Gold 6148, 2.4 GHz, 40 cores/node
Memory: 96 GB/node, BW 255 GB/s/node
Interconnect: Infiniband EDR, Fat tree

ー Storage

Online storage (OFS): 5PB, FEFS
Hierarchical Storage (HSM): 300 TB (cache disks), 7.9 PB (tape)

The Number of Supercomputer Users

Used core time for each field

Chemistry - organic chemistry: 5.29%
Chemistry - others: 0.09%
Chemistry - physical chemsitry: 17.61%
Engineering - applied mathematics: 0.03%
Engineering - information engineering: 0.20%
Engineering - mechanical engineering: 1.00%
Engineering - others: 2.21%
Life Science - biochemistry/molecular biology: 5.18%
Life Science - bioinformatics: 9.11%
Life Science - biophysics: 18.55%
Life Science - brain science: 0.58%
Life Science - others: 0.01%
Others: 0.11%
Physics - astronomy/cosmology: 1.81%
Physics - condensed matter physics: 11.37%
Physics - others: 0.00%
Physics - particle and nuclear physics: 26.86%

HOKUSAI SailingShip (SS)

ー CPU farm: 440 nodes, 1.62 PFlops

CPU: Xeon Platinum 8260, 2.4 GHz, 48 cores/node
Memory: 384 GB/node, 281 GB/s/node
Disk(SSD): 1.92TB
Interconnect: 10 GBBASE-T x 2

ー Data farm

30PB, Lustre files system
NFS/CIFS/S3 gateway

ー Private Infrastructure as a Service (PIaaS)

Mirantis Cloud Platform (MCP)
ーCommercial distribution of OpenStack that can provide VM, container and bare metal environment

HOKUSAI Team <isys-hpc@ml.riken.jp>

Head Office for Information Systems and Cybersecurity

http://i.riken.jp/en/

Knowledge Organization and Open Science

Collect, organize, integrate and promote the use of various RIKEN’s research data as an Open Science

RIKEN is the largest Japanese comprehensive natural science institute including various research centers and many small laboratories having about 3,000 researchers.

Contribution to maximizing research results through research data knowledge organization based on ontology and metadata research
Realising description, collection and publication regardless of the research fields. Creating cross-field integration formation knowledge
Practicing metadata-driven research using the integrated knowledge. Leading open science and contribute to create innovation

cf. RIKEN MetaDatabase: An open-life-science platform that provides multi-field integrated knowledge as a set of metadata (https://metadb.riken.jp)

The research data knowledge organization based on ontology and metadata research

Achieve unified data management
regardless of research field by
developing a metadata-driven repository

RIKEN Research Data Management Service

We develop and operate a data management service that accumulates a variety and large amount of research data produced by RIKEN and supports data processing and utilizing, based on an appropriate open/close strategy. such as:

Management methods according to individual policies for each research institution and researcher
Jointing methods with supercomputers, storages, clouds, etc.
Efficient and extensible methods to use storage, database, and cloud technique
Optimization of operation and user support, and, for that purpose, development of usability improving tools

Our service plans to accept users outside of RIKEN via GakuNin(academic authentication federation) and also accept other organization that has jointed research with RIKEN.

DMS: Data Management System Unit

DKO: Data Knowledge Organization Unit

figure of Design of RIKEN Research Data Management Service — Fig1: Design of RIKEN Research Data Management Service

figure of Coordination of other infrastructure — Fig2: Coordination of other infrastructure

Hideyuki Jitsumoto, Data Management System Unit (DMS)

https://www.riken.jp/en/research/labs/isc/data_mange_sys/index.html

Norio Kobayashi, Data Knowledge Organization Unit (DKO)

https://www.riken.jp/en/research/labs/isc/data_knowl_org/index.html