System Design · NVIDIA · Hard
Requirements Create a centralized control-plane service for an HPC software system. Functional requirements: Keep the up-to-date status of a large fleet of hosts within the control plane. Allow hosts to report their health, availability, and resource information to that service. Coordinate job placement decisions, without requiring an in-depth scheduler design. Persist host and job state in a database using a well-defined schema. Remain correct under heavy concurrent updates…
Checking your access…