System Design AI
Explore designs

Design preview

Design a GPU Cluster Scheduler for Training Jobs

Thousands of GPUs shared by many teams: gang admission, topology-aware placement, quotas with lending, checkpointed preemption and failure recovery: one-hour boards for junior, senior and staff, with the theory behind them.

ai-infragpuschedulingkubernetesawsinterview-board

Shared by System Design AIOfficial

Outline of the design layout

Explore the complete design

Open the diagram and design notes, discuss trade-offs with the community, or make a private copy to build with Coach.

Checking your sign-in…