Access to the supercomputers

General principles

The access to tier-0 and tier-1 systems is subject to the following general principles since the HBP currently does not own supercomputer or data center resources.

  • No large-scale in-kind contributions of resources are foreseen in SGA1 (RUP in-kind contributions are not transferable to SGA1).
  • For SGA1, available resources are PRACE/European or national resources. In this context, applications are subject to the rules of the providers of these resources.
  • Preparatory and development resources are provided under separate conditions, as explained in more detail below.

Large-scale Projects

1. European resources (PRACE; called Tier-0) 

(The following is subject to a signature of PRACE 2.0 foreseen in September 2016).
In the coming second phase of PRACE operations (PRACE 2.0), PIs from hosting members* (HM) can apply starting with call 14. These resources are open to all sciences in a competitive manner. Applications will underlie the peer review process of PRACE.
In general, 30 % of the available resources of the contributing HMs will be given to PRACE for the HM PIs.
In general, PIs from contributing general partners* (CGP) will have access to apply for up to 10 % of resources provided by the HMs.
*PRACE hosting members are committed to fund and deliver large PRACE RI computing and data management resources. Under PRACE 2.0 there will be five HM: France, Germany, Italy, Spain and Switzerland. Contributing general partners are other countries that do not contribute hardware resources but e.g. support. All 27 EU member states are members of PRACE.
PRACE 14th Call for Proposals for Project Access is open until 21 November 2016.

2. National resources (called Tier-1)

In general, PIs being citizen of the country that offers resources can apply. Applications are under the national peer review rules.
Some countries accept large-scale applications PIs from foreign countries:
  • Jülich Supercomputing Centre (JUELICH) will consider those applications to a certain extent. They will be subject to the transparent scientific peer review under the John von Neumann Institute for Computing. There is no restriction applied for Co-PIs and collaborators.
  • The national resources provided by the User Lab at Swiss National Supercomputing Centre (CSCS) are open to all scientists, irrespective of their nationality, provided the projects comply with US Export Control. These resources are subject to a competitive allocation process with transparent peer review.
  • For national calls at BSC it is recommended but not mandatory to have a Spanish collaborator in the project.
  • The national calls in Italy are open to PIs affiliated to an Italian institution, while no restriction is applied for the Co-PI and collaborators. The Italian resources are provided subject to a competitive allocation process with transparent peer review. However, Cineca additionally provides moderate amounts of computing time to HBP scientists.

The following national calls are available for the systems integrated into the HPAC Platform:

3. Future HBP assigned resources

In Europe, PRACE is acknowledged as the European HPC infrastructure by the European Commission and all 27 member countries of PRACE. Funding HBP supercomputing resources, the EC expects all large-scale allocations to be provided through PRACE. Therefore, future HBP resources provided by the EC will undergo the PRACE peer review. The full amount of these resources is foreseen to be provided to the HBP.

Preparatory and Development Projects

Furthermore, the members of the HPAC platform, BSC, CINECA, CSCS and JSC are providing preparatory and development access. The former typically are intended to prepare larger allocation projects and range from 5,000 to 100,000 core hours, depending on the systems. Development projects serve major algorithm and code development purposes. Preparatory and development allocation requests are handled locally by these sites and typically require only technical reviews. The size of development projects at CSCS can be as large as one million core hours.