Open Call: Advanced Computing Projects Competition (5th edition) – Portugal

Objectives

• The consolidation and strengthening of the National Scientific and Technological System (SCTN) are priorities of Portugal’s science and technology policy. These priorities aim to contribute to the national and international competitiveness of science and technology, as well as their role in innovation and knowledge transfer. Furthermore, they seek to contribute to the fulfillment of global aspirations defined in the United Nations Sustainable Development Goals (SDGs). In this context, promoting and enhancing the skills of scientific and technological institutions through the participation of their teams in advanced computing projects is of particular importance.
• Through the National Advanced Computing Network (RNCA), the FCT seeks to aggregate national advanced computing resources, fostering cooperation among the various involved centers and developing national and international partnerships with other entities.

Form of support

• The support granted under this call is exclusively in the form of allocated usage time of advanced computational resources, without providing any funding of any kind or human resources to develop or support software applications.

Access Types

• The call includes four access types:
  • A0 – Experimental Access:
    • This type of access is recommended for scientific or innovation projects whose team lacks prior experience in advanced computing or does not have a history of using computational resources. It is intended for experimentation, testing, and pilot access to the platforms. Applications for this type of access will be subject to administrative validation and technical adequacy assessment.
  • A1 – Development Access:
    • This type of access is recommended for conducting software performance tests, code optimization, scalability testing, benchmarking, refactoring, and small-scale projects. Applications for this type of access will be subject to administrative validation and technical adequacy assessment.
  • A2 – Regular Access:
    • This type of access is intended for the use of HPC, AI, and/or Cloud resources on the Deucalion, Cirrus, and/or Stratus platforms and is recommended for scientific or innovation projects with a team that has prior experience. To demonstrate the adequate scalability of access requests, the operational teams of the platforms involved in this call may require a prior A0 or A1 access. Applications for this type of access will be subject to administrative validation, scientific merit assessment, and technical adequacy validation.
  • A3 – Larger Scale Access:
    • This type of access is exclusively intended for the use of high volumes of HPC and/or AI resources on the Deucalion or MareNostrum 5 platforms and is recommended for scientific or innovation projects with a team that has prior experience in HPC and/or AI. To demonstrate the adequate scalability of access requests, the operational teams of the platforms involved in this call may require prior access of A0, A1, or A2. Applications for this type of access will be subject to administrative validation, scientific merit assessment, and technical adequacy validation.

Computational Models

• This call aims to allocate computational resources to projects in all scientific and innovation domains, following international technological standards. The following computational models are available:
  • High Performance Computing (HPC)
    • For the purposes of this call, each HPC 1 architecture consists of the following elements:
      • A set of compute nodes that operate together and are temporarily dedicated to a single application. Collectively, these nodes can execute at least 40 x 10¹² [ 2 ] floating-point operations per second, which are highly interdependent, executed on general-purpose, non-specialized microprocessors.
      • A file system accessible from each compute node with a shared throughput of at least 40 Gbps, supporting multiple simultaneous access streams per node.
    • The compute nodes are typically managed by a batch system such as Slurm or a similar system. HPC systems are generally accessed via SSH to one or more entry nodes, from which jobs can be submitted to the batch system. In the context of this notice, teams with or without prior experience are also allowed to request resources for visualization (e.g., GPUs dedicated to this purpose).
  • Computing for Artificial Intelligence (AI)
    • This model aims to support research and development projects that use artificial intelligence tools and data analysis algorithms in areas such as Natural Language Processing (Natural Language Understanding), Ethical Artificial Intelligence, or other related fields.
    • Additionally, HPC resources (CPU, GPU, and storage) can be made available to compute and store data, supporting the development, testing, and implementation of various applications in the fields of Artificial Intelligence, Data Science, and Big Data analysis.
  • Scientific Cloud Computing (Cloud)
    • In the context of this notice, each Cloud architecture is composed of the following elements:
      • A set of compute nodes shared by multiple users and applications, provided in a self-service system with maximum resource usage quotas through a virtualization software layer in IaaS (Infrastructure as a Service) cloud computing.
      • The virtual servers (VM – Virtual Machines) provided access virtual disks either through local devices or by mounting a remote file system.
    • The creation of VMs can be done via a web dashboard, command-line tools, or APIs. The service is based on OpenStack and is designed for scientific data processing in a cloud computing environment. This model allows the user to fully define VMs, including the Linux operating system, hardware, and software configuration, offering great flexibility for setting up and using resources for computational tasks.

Target Beneficiaries

• The allocation of computational resources can be made in the form of Individual or Institutional Support, that is, to individuals or institutions, either individually or in copromotion, as mentioned in Articles 3, 4, and 6 of the Advanced Computing Projects Regulation.
• Regarding applications from companies as beneficiaries, advanced computing projects must:
  • Occur within the scope of pre-competitive research and innovation, where the goods or services resulting from that research or innovation have not yet been assigned commercial value.
  • Not exceed, in total, 50% of the computational resources to be allocated in this call for all applications of this type.

Eligibility Criteria

• Applications are accepted either individually or in co-promotion, under the Individual or Institutional Support modalities:
  • To one or more computational models;
  • To one or more platforms for the same project;
  • To one or more distinct access types (A0, A1, A2, or A3)
• In the case of applications under access types A2 and A3:
  • An eligibility criterion for the HPC computational model is the submission of a scalability graph of the software to be used, obtained from real or estimated data. It is also recommended to present prior experience, particularly in the use of this computational model or in previous advanced computing projects (e.g., in RNCA, PRACE, EuroHPC, etc.).
  • Eligibility for the Cloud computational model for commercial and/or profit-oriented entities is limited to the availability of physical computational resources after allocation to non-profit or non-commercial entities and is also subject to any costs indicated in the acceptance term of the computational project.
• The Principal Investigator (PI) of the project:
  • Must update and provide their CienciaVitae, associated with CienciaID, to the FCT at the time of application;
  • In access types A2 and A3, the Principal Investigator (PI) must identify a co-investigator responsible for the project, referred to as the Co-Principal Investigator (co-PI), who will substitute the PI in their absences, unavailability, and impediments.

For more information, visit FCT.