Re-Sourcing? Optimised Efficiency in Academic Networks via Shared Computational Resources
Wednesday, June 24, 2026 3:45 PM to 5:15 PM · 1 hr. 30 min. (Europe/Berlin)
Foyer D-G - 2nd Floor
Project Poster
Composable Disaggregated InfrastructureHeterogeneous System ArchitecturesHPC in the Cloud and HPC ContainersNetworking and InterconnectsResource Management and Scheduling
Information
Poster is on display.
Access to high-performance GPU resources remains as a persistent bottleneck for academic research, despite substantial investment in local HPC infrastructure. Researchers frequently face long queue times during periods of peak demand, while compatible capacity remains idle elsewhere in the academic ecosystem. This fragmentation leads to underutilisation of existing resources and constrains research throughput.
Prior efforts, such as the National Research Platform in the United States, have demonstrated that coordinated access to academic compute is possible within bounded contexts. However, such approaches are typically programme-specific or national in scope, and no widely adopted, general-purpose global coordination mechanism for academic compute currently exists. In addition, there is no shared agreement on how coordination across institutions should be defined, implemented, and governed at a global level.
This work investigates the hypothesis that improved workload coordination, rather than additional compute capacity alone, can increase effective utilisation of existing academic infrastructure. It introduces the Academic Compute Network as a thin coordination and distribution network that enables workloads to move to available capacity across heterogeneous, locally managed environments, while preserving institutional sovereignty and existing HPC and cloud systems.
Access to high-performance GPU resources remains as a persistent bottleneck for academic research, despite substantial investment in local HPC infrastructure. Researchers frequently face long queue times during periods of peak demand, while compatible capacity remains idle elsewhere in the academic ecosystem. This fragmentation leads to underutilisation of existing resources and constrains research throughput.
Prior efforts, such as the National Research Platform in the United States, have demonstrated that coordinated access to academic compute is possible within bounded contexts. However, such approaches are typically programme-specific or national in scope, and no widely adopted, general-purpose global coordination mechanism for academic compute currently exists. In addition, there is no shared agreement on how coordination across institutions should be defined, implemented, and governed at a global level.
This work investigates the hypothesis that improved workload coordination, rather than additional compute capacity alone, can increase effective utilisation of existing academic infrastructure. It introduces the Academic Compute Network as a thin coordination and distribution network that enables workloads to move to available capacity across heterogeneous, locally managed environments, while preserving institutional sovereignty and existing HPC and cloud systems.
Format
on-demandon-site