How MEGWARE Came to AlmaLinux for HPC
summary
MEGWARE explains why AlmaLinux replaces CentOS in long‑lived HPC clusters, citing InfiniBand history, two‑week reinstall cost for major upgrades, and use of local mirrors to keep node images identical. Production facts: AlmaLinux 8.10 running on a master node, a fully liquid‑cooled GPU cluster rated at 52 petaflops, and a planned GP200 NVL4 project with 1,000+ GPUs and a €650M budget. Notes include AMD LINPACK regression debugging, an emerging AlmaLinux HPC repo, and silver sponsorship to accelerate support and hardware enablement.
AI-written summary · about
chapters (15)
- 0:00Introduction to MEGWARE
- 2:16Why RPM-based distributions
- 3:43CentOS discontinuation impact
- 5:47First AlmaLinux deployment
- 7:34Post-Red Hat source code announcement
- 8:39Becoming AlmaLinux sponsor
- 10:02Customer support concerns
- 11:27Community and upgrade speed
- 12:29Extended hardware support
- 13:52Additional software repositories
- 14:54CUDA and Nvidia support
- 15:15CERN’s adoption impact
- 15:57Top 500 list achievements
- 17:04Latest AlmaLinux projects
- 19:13Q&A
related talks
16mAlmaLinux in brief
12mAlmaLinux the First Three Years and Beyond
33mThe Future of AlmaLinux: Collaborating on Development and Adoption from Japan
17mAlmaLinux and the Open Source way
32mALESCo, corporate contribution, and the way forward
51mGoing Linux on Massive Multicore
32mAlmaLinux in Japan: Progress and Achievements Over the Past Year
41mPodman containers and HPC