Rack-scale GPUs and scheduling squeeze

NVIDIA is framing its Blackwell systems as rack-scale supercomputers that need topology-aware scheduling, and it has introduced Mission Control software to bridge that hardware with AI schedulers. At the same time, Nvidia's acquisition of SchedMD — the maker of the Slurm scheduler — is drawing scrutiny because it concentrates both chips and scheduling roadmap in one vendor, which could create lock-in for mixed‑vendor clusters. (developer.nvidia.com) (blockchain.news) (infoworld.com)