Workshop: Sovereign AI Supercomputing Cloud Workshop
Authors: Helena Vela Beltran (Do IT Now)
Abstract: Managing workloads across rapidly diversifying hardware—spanning x86, Arm, and emerging RISC-V CPUs alongside competing NVIDIA and AMD GPU architectures—demands a software infrastructure that is genuinely independent, vendor-neutral, and resilient against ecosystem lock-in.
This talk introduces the European Environment for Scientific Software Installations (EESSI), a community-governed software stack designed to provide an autonomous, transparent foundation for heterogeneous scientific computing. Supported through the EuroHPC Joint Undertaking via the MultiXscale Centre-of-Excellence, EESSI delivers an on-demand, streaming distribution architecture (powered by CernVM-FS) that decouples scientific applications from specific host operating systems and single-vendor software ecosystems. By dynamically serving microarchitecture-optimized binaries across diverse CPU targets and GPU acceleration platforms without compromising performance, EESSI eliminates single-site redundancy and prevents vendor-driven stack fragmentation.
Ultimately, we demonstrate how establishing a shared, institutionally neutral software ecosystem ensures long-term scientific reproducibility, and returns software governance to the open scientific community.
Back to Sovereign AI Supercomputing Cloud Workshop Archive Listing
Back to Full Workshop Archive Listing