Scope

The objective of Computational Simulation Centre (CSC) is to provide and exploit a supercomputer for large scale simulation activities to analyse experimental data on fusion plasmas, prepare scenarios for ITER operation, predict the performance of ITER facilities, and contribute to DEMO design physics basis and BA activities.

Schedule

The CSC activity has been very successfully completed in May 2017 in accordance with the IFERC project plan and yearly work programmes based on the procurement arrangements.

To effectively coordinate the activity, it is broken down into two phases, followed by a dismantling:　

The CSC activity has been very successfully completed in May 2017 in accordance with the IFERC project plan and yearly work programmes based on the procurement arrangements.

 Phase One: Computer Selection and Procurement Phase (Jul. 2007 – Dec. 2011).
 Phase Two: Computer Operation Phase (Jan. 2012 – Dec. 2016).
 Decommission: Dismantling of computer equipment (beginning of 2017).

Phase One :

The goal of Phase One is to set up the supercomputer and the associated peripheral equipment (including a data storage system and a visualisation system) in the CSC&REC Building located in the Rokkasho BA site and to commission the equipment in order to get ready to start Phase Two (operation).

It is important to involve early in the process the fusion community, in order to ensure that the appropriate codes are developed for the simulation and modelling of ITER, of the other fusion experiments, and for the design of DEMO and future fusion power plants.

To take into account the needs of the users, Special Working Group (SWG-1) was set up to recommend target technical specifications for the IT equipment and associated services.

The SWG-1 started in 2008 with the missions of defining the areas of applications, drawing up the actual minimum expected performance, assisting the procuring Implementing Agency (IA) in a market research, selecting benchmark codes, assessing the performance of the benchmark codes on candidate High Performance Computing (HPC) systems, reviewing the technical specifications for the HPC formulated by the procuring IA, defining user requirements for maintenance, operation and technical support, and monitoring the acceptance tests and the early operating phase of the HPC system. In 2012, considering the usefulness of the contributions of SWG-1 during the early operation phase, it was proposed to extend the mission of SWG1 to continue to represent the needs of the users of CSC in the EU and JA fusion communities.

The Procurement Arrangement for the supply of a supercomputer and peripheral equipment was signed in 2010, and the contracts for the supply of the equipment, and operation and maintenance were signed in March 2011. Other Procurement Arrangements were signed to take care of the various interfaces. The main task of the EU-IA is to procure the supercomputer, the peripheral equipment and the operation and maintenance associated services. The main task of JA-IA is to prepare interfaces for the installation of the equipment and to contribute to the seamless integration of the IT equipment and services in the International Fusion Energy Research Centre, in particular by providing support for the interface with the users. For this purpose, the two agencies have conducted intensive preparation work.

In order to prepare the operation phase, the second Special Working Group (SWG-2) established in 2011 the project selection procedures, user utilisation rules, and the role of Standing Committee to preside the process. The Standing Committee was established accordingly in 2011. Its role includes promoting the lighthouse projects, making calls for proposals, selecting the projects from the proposals by the peer review, allocating the computational resources to the selected projects, evaluating the performed projects, making a written summary report of the call, and encouraging the dissemination of results.

Fusion for Energy took ownership of the super computer systems upon the completion of the "Ready to run acceptance tests" in December 2011 as defined in the Procurement Plan of Procurement Arrangement for the supply of the supercomputer and peripheral equipment for the IFERC project (CSC sub project). The supercomputer is named Helios and it has a peak performance of more than 1.5 Petaflop/s.

Phase Two :

The "ready to run" acceptance of Helios marked the end of Phase One and the beginning of phase Two.

The goal of Phase Two is the effective operation of Helios and its sharing by the users. Phase Two started on schedule in January 2012.

In the first quarter of 2012, the operation of Helios has been dedicated to four "lighthouse projects". Since then and until the end of 2016, Helios is running in normal operation as defined by SWG-2. In 2013, under the agreement of SWG-1, an increase of the memory capacity of the login nodes and various small enhancements have taken place to improve the operation of Helios.

In the first quarter of 2014, new nodes containing Intel Xeon Phi coprocessors were installed, increasing the total computing capacity by more than 200 Teraflop/s peak. Another upgrade was done in the first quarter of 2016 with additional nodes equipped with NVidia K80 GPU (35 Teraflop/s peak in total). The purpose of these two complementary upgrades is to help the fusion community to prepare for the next generation of supercomputers. These two enhancements are under the management of both IAs.

The dismantling of Helios is planned for the 1st semester of 2017.

Structure

Activities

Operations

The Helios supercomputer has finished its operation in December 2016.

Deadline for Project Report for the 5th CSC Cycle

Deadline: 31 January 2017

Deadline for Project Report for the 4th CSC Cycle

Deadline: 29 January 2016

Helios Expanded with new cluster

Additional nodes equipped with NVidia K80 GPU have been installed.

The 5th CSC Project Cycle is in progress

Start Date: 17 November 2015
End Date: 27 December 2016

Deadline for Project Report for the 3rd CSC Cycle

Deadline: 31 January 2015

The 4th CSC Project Cycle has Started

Start Date: 2 December 2014
End Date: 17 November 2015

The HELIOS supercomputer ranked as #38

The HELIOS supercomputer system at IFERC-CSC has been ranked as #38 with 1.237 Peta Flop/s (more than one thousand trillion calculations per second) in the most recent world's TOP500 list [http://www.top500.org/lists/2014/11].

Upgrade to Intel Xeon Phi Native Mode Operation

The operation of Intel Xeon Phi native mode began concurrently with the beginning of the 4th CSC project cycle. With 180 MIC nodes, each comprising two Intel Sandy Bridge processors and two Intel Xeon Phi co-processors, a total of 24480 cores (21600 of them on the co-processors) is available for accelerating applications.

HELIOS Expanded with New Computer Racks

An additional rack was installed and new nodes were added to the production. With the enhancement, a total of 4500 nodes (72000 cores) is available for use.

Deadline for Call for Proposals for the 4th Cycle from
December 2014 to November 2015

Deadline: 30 July 2014

* We have closed the call for proposals for the use of the Helios supercomputer in the 4th cycle. Thank you to all those who submitted proposals.

Reports of the previous calls (1st - 3rd) are available here:

Report of Standing Committee on the 3rd Call [PDF: 175KB]
Report of Standing Committee on the 2nd Call [PDF: 131KB]
Report of Standing Committee on the 1st Call [PDF: 130KB]

Meetings

Achievements

Helios supercomputer was operated with a very high availability since the beginning of the operation and with a very high usage rate by the fusion community:

A large number of accumulated peer-reviewed papers (639 as of February 2017) were published in major scientific journals with a wide range of activities from plasma physics to reactor materials and technology:

This page chronologically lists publications that Helios users produced as a result of using the CSC facility. Reported publications are classified in 9 categories:
   Linear, non linear and/or extended MHD
   Fast particle physics
   Plasma turbulence and related transport processes
   Edge physics
   Heating and current drive
   Integrated modeling
   Reactor materials
   Reactor technology
   Computational science

★ 2016

★ 2015

★ 2014

★ 2013

★ 2012

 

Systems

System Specifications

Processor Specifications

Network and Storage Infrastructure

For Users

Events

Contact us

We can be reached

At Rokkasho-site:

International Fusion Energy Research Centre (IFERC)
Computational Simulation Centre (CSC)
2-166 Oaza-Obuchi-Aza-Omotedate
Rokkasho-mura, Kamikita-gun, Aomori 039-3212
Japan

By E-mail:

Contact

IFERC HPC follow-up Working Group

Purpose

An “IFERC HPC follow-up working group” was set-up in 2017 in order 1) to address the topics to share experience and best practices in the design and operation of HPC centres for fusion users and in the usage of such centres by fusion users in Europe and in Japan, 2) to prepare/monitor “Joint EU-JA HPC simulation projects” from April 2019 to March 2020 by providing computer resource equivalent to 0.2 PFlops Linpack from both EU (A3 and A2 partitions of MARCONI) and JA (JFRS-1), and 3) to consider the possibility and to provide input for the joint procurement of a supercomputer in the context of BA Phase II based on the expected lifetime of the current systems in EU and in JA.

Activities

Joint EU-JA HPC simulation projects

Purpose

The purpose of the “Joint EU-JA HPC simulation projects” is to foster the collaborative joint simulation projects between EU and JA researchers in the magnetic fusion energy research and development including reactor materials and technology as wider as possible, by providing a fraction of the supercomputer funded by EUROfusion for EUROfusion users (currently Marconi-Fusion located in CINECA) and a similar fraction of the new supercomputer funded by QST for Japanese users (currently JFRS-1 located in Rokkasho) from April 2019 to March 2020.

Available Resources

The total amount of available node hours for this call, in Marconi A3 (conventional nodes) and in Marconi A2 (accelerated nodes) are 916,000 node-hours and 412,000 node-hours, respectively. The total amount of available node hours in JFRS-1 (conventional nodes) are 858,480 node-hours.

Selection Process

In order to handle those joint simulation projects, the joint allocation committee was set up in December 2018, whose role is to submit the call for proposal, to select the projects, to allocate the CPU time to the selected project and to evaluate the users reports as was described in Terms of Reference (ToR).
The joint allocation committee launched “Call for Proposals” with the deadline of March 8, and in total 10 proposals were received. The 10 proposals are categorized into 2 for Turbulence and transport, 4 for Energetic particles, 1 for MHD and Energetic particles, 2 for Reactor Materials, 1 for Heating technology (NBI).
The joint allocation committee implemented the peer review to rank the proposals based on the 4 criteria described in the "Call for proposals", and the peer review process ended on 19 March. Then, the committee selected the top 5 out of 10 proposals based on the ranking by the peer review, taking into account 1) fostering the collaborative joint simulation projects between EU and JA researchers in the magnetic fusion energy research and development including reactor materials and technology as wider as possible suggested in the "Call for Proposals", and 2) assigning a sufficient CPU time for researchers of the selected projects to implement the production runs leading to achievements to be reported.
The selected “Joint EU-JA HPC simulation projects” are summarized below: