Description: By the end of the action, an effective application support team network across EuroHPC stakeholders (primarily Hosting Entities) will have been established providing services to European HPC users from public and private sector including SMEs. This network will provide a knowledge hub for HPC offering a comprehensive support service from porting, optimisation and execution of key applications as well as training and development of specialised cross-domain support material that will facilitate EuroHPC system exploitation and delivery of excellent scientific results.
Basic Information
Reference Number: AS-023-8455
Project Type: DIGITAL-JU-SIMPLE DIGITAL JU Simple Grants
Type of MGA: DIGITAL Action Grant Budget-Based [DIGITAL-AG]
Details
Location: –
Work programme part: Application Support in High Performance Computing
Dates
Publication: 10.09.23
Closing Date: 10.03.24
Contact Information
Description
xpectedOutcome:
By the end of the action, an effective application support team network across EuroHPC stakeholders (primarily Hosting Entities) will have been established providing services to European HPC users from public and private sector including SMEs. This network will provide a knowledge hub for HPC offering a comprehensive support service from porting, optimisation and execution of key applications as well as training and development of specialised cross-domain support material that will facilitate EuroHPC system exploitation and delivery of excellent scientific results.
In the course of the action, the Application Support Teams will have demonstrated their significant impact on the application enabling with success stories within the European HPC ecosystem supported by quantitative key performance indicators measuring the specific impact of individual applications, scientific domains and target systems, with respect to the baseline established.Objective:
The call aims for the establishment and four-year operation of a distributed European-wide high- performance computing application support service. The selected project should bring together Application Support Teams (AST) established (primarily) in current or future EuroHPC Hosting Entities, operating the EuroHPC supercomputers. ASTs will work closely with the EuroHPC JU Peer Review Office managing the evaluation and allocation of access time to European HPC supercomputers as the applications and respective user teams to be supported will be primarily selected by peer-review process managed by the EuroHPC JU. Through this process ASTs are expected to provide their services to a selected number of flagship/lighthouse which have been allocated time in the context of the EuroHPC JU Extreme Scale and Regular Access calls.Scope:
ASTs should offer a broad range of support services aiming primarily on Level 2 and 3 application support encompassing at least the following services:
Application porting and optimization
- Code enabling and scaling
- Elaboration of performance analysis and benchmarks for selected applications
- R&D on code refactoring
- Short to medium term support of selected applications/user groups (3-18 months)
- Domain independent, architecture specific trainings (targeting the EuroHPC supercomputers)
- Development of best practice guides for using the EuroHPC supercomputers.
- Collaboration with equivalent groups at European and International level on technical matters related to supporting pre/exascale application work.
- Contribute to the development and improvement of the European HPC Application Support service by sharing lessons learned and providing feedback to the EuroHPC JU on user requirements and service gaps.
- Customised support to scientific and research community from public and private research institutes and academia, industry and SMEs.
The project should establish a coordination layer between the participating teams, allowing exchange of knowledge, information, as well as harmonization of a range of services and service level provided. The size of the selected ASTs should be proportional to the size of the local system supported and is expected to be in the range of 4-6 experts for pre-exascale level systems and beyond, and in the range of 2-4 experts for petascale and mid-range systems.
