Coordinators: Andrew Katumba (Makerere University) , Tim Kamanu (University of Nairobi)
The overall goal of this project is to establish a federated HPC infrastructure to support large scale modelling, AI workloads, and other compute and storage intensive tasks. The initial view is rather than have one centralized HPC system, is to have at the very least, 3 HPC needs hosted at a Core AI partner institution in each of the regions. The infrastructure would be accessible to all consortium members and their affiliates and also serve as hub for national and regional universities. The hub would be organized with the typical HPC Tiered system. This would for example allow for other non-academic, non-consortium partners to access the resources at a premium level (Tier 1). Researchers with project budget could also temporarily expand their resource quotas by purchasing extra credits. These two are key to sustainability of the project. For this infrastructure project, we propose Stellenbosch, Makerere and a partner in Nigeria (to be determined) as a host to establish regional presence. All tier 2 facilities would be accessible with no cost to all partners.
Swedish (and other EU partners) would provide training, support, procurement and operational guidance for the project. It is also conceivable that initially access to some of the HPC infrastructure available at Uppsala to core AI members to offload their compute needs for ongoing project activities.
Consultative meetings with ICTP and existing HPC systems in Sweden or EU would help to better guide the conceptualization. For example the ICTP HPC graduate programme could be tagged for initial talent. Another natural connection can be established with the newly created ICTP International Consortium for Scientific Computing