Go with the (Energy) Flow: Adaptive Capacity Computing

With the ever growing size and complexity of its systems and applications, HPC has become a significant consumer of electrical energy, with Exascale sites having power ratings comparable to very large industrial installations. The rapid switch towards inherently variable Green energy sources like wind and solar impacts HPC centres and their users in a fundamental way: the grid power required for full system operation might not be available 24x7, and even if it is, prices will vary according to actual availability and consumption. Under such circumstances, HPC centres can no longer count to be able to run their systems with full capacity – either forced by energy availability or costs, or on their own choosing to help stabilize the Grid. “Adaptive Capacity Computing” touches on system architecture, hardware, scheduling and resource management, programming models and applications with the objective of enabling future HPC centers to react gracefully to varying power profiles, achieving optimal through put and avoiding loss of computational state wherever possible. This BoF continues the discussion started at SC23 on approaches for the HPC community at large to provide adaptive capacity computing, and how to minimise the challenges created for HPC application developers and users.