Science Data Challenge 3

JPSRC is in a transition phase from the prototyping to the SRCNet v0.1, so that a major upgrade of the JP-SRC system with some maintenance is planned during the SDC3b season.

Resources per team

• Workstations or virtual machines are assigned per team

Resource Access

• SSH access with the ssh-key and the password, or Web access with the local account or through eduGain. VPN may be asked in the former case. Users accessing from outside of Japan may need to undergo the relevant regal security checks.

Data cube access

• SDC3 data cube will be stored at local storage, to which users/teams can access

Resource management

• Users/teams will share the resources with other users/teams. 

Software management

• Users/teams can install their own software/docker where possible but may need to request installations by JP-SRC superusers.

Documentation

• A brief instruction will be provided in English and Japanese languages.

Support

• Support from JP-SRC will be provided via E-mail and Slack

Resource location

• Tokyo, Japan

Overview

• JP-SRC is an inter-university cooperation led by the SKA1 Japan Promotion Group (SKAJ) of the National Astronomical Observatory of Japan (NAOJ). JP-SRC is currently operating prototype servers which are utilised for data analysis for SKA precursors and pathfinders. From SDC3 JP-SRC offers these servers for science data challenges as well.

• It is a transition phase from the prototyping to the SRCNet v0.1, so that a major upgrade of the JP-SRC system with some maintenance is planned during the SDC3b season. The JP-SRC will be a three-layer system; CANFAR science platform running on the Kubernetes clusters under the OpenStack virtual machines. Until the system is commissioned and opened, JP-SRC will provide users/teams individual workstations of JP-SRC.

Technical specifications

• JP-SRC contains heterogeneous workstations, with (in total):

    •152 cores (6.41 TFlops FP64)

    • 6 GPUs (242.57 TFlops FP32, 13.98 TFlops FP64)

    • 1 NEC SX-Aurora TSUBASA vector engine  (2.1 TFlops FP64)

    • 1.86 TB DRAM memory, 3.71 TB M.2/SSD

    • 338 TB HDD storage

• Network speed is typically 1 Gbps

Per user resource

• Resource per user/team will be determined by JP-SRC by taking into account the number of users(teams) and the available resources.

• JP-SRC will allocate reasonable resources as requested for the analysis, first-come-first-served basis.

Software installed

• Linux OS. A popular OS is Ubuntu 22.04. Docker and Singularity will be supported.

• Participants can install their own software where possible but may need to request installations to JP-SRC superusers.

Volume of Resource

• The volume of resources is ultimately limited by the physical capacity of the assigned workstations.

• Due to the limited resources, the number of users will be limited up to ~10.

GPUs if any

• 2 x NVIDIA Quadro GP100 16GB

• 2 x NVIDIA GeForce RTX3090 Turbo 24GB

• 2 x NVIDIA Quadro A6000 48GB

Setting up and Account

In order to set up an account please contact skaj-src-inquiry@ml.nao.ac.jp.

Logging in

• SSH access (username, ssh-key, and password) to login to the assigned workstations. VPN may be asked for some workstations.

• Web access (username and password) using the local account or eduGain.

• Users accessing from outside of Japan may need to undergo the relevant regal security checks.

• A few team members, who will actually analyse the data, per team will be awarded access permission

How to run a workflow

• Users are free to run their workflows to their assigned workstations or virtual machines.

• Command-line only or web-based environments such as Jupyter notebook will be supported later.

• 
  

Accessing the data cube

• Primary data cube provided by SKAO for SDC3 will be stored/cloned by JP-SRC at local storage, to which a team can access.

• User can make multiple copies on the assigned storage.

Software management

• User will install their own software where possible but may need to request installations by JP-SRC superusers.

Containerisation

• Users can install and use docker or singularity to run containers on their workstations.

Documentation

• A brief instruction will be provided in English and Japanese languages.

• The JP-SRC coordinator will manually manage and monitor the resources.

• Users/teams will share the resources with other users/teams.

• Users/teams affiliated with Japanese research institutions will be prioritised, if there are too much requests.

• Support from JP-SRC will be provided via email and Slack.

• JP-SRC is funded by SKAJ and is cooperated by Kumamoto University and Nagoya University.