Which cluster suits MIG best?
Why A is the answer
Many small jobs that underuse a full GPU
Small jobs fit in partitions.
In detail
- Small jobs fit into partitions.
- Large training jobs need full GPUs.
Use case
An illustrative scenario, not a real incident.
A cluster running many small inference jobs suits MIG, as each fits in a partition.
See it in different ways
flowchart LR A["Enable MIG mode"] --> B["Create GPU instances"] B --> C["Create compute instances"] C --> D["Assign to workloads"]
mindmap
root((Sharing))
Run:ai
Project quotas
Over quota, then preempted
Slurm
Fairshare
MIG
GPU instances
Hardware isolationflowchart TD
Q{"What are the jobs like?"}
Q -->|"Small"| A0["Good fit"]
Q -->|"Need a full GPU"| A1["Poor fit"]Key point: MIG suits many small jobs.
Why the other options are wrong
- B. One huge job that needs every GPU
- That job needs whole GPUs.
- C. A cluster with no GPU workloads
- There is nothing to share.
- D. One job per GPU that runs at full load
- Those GPUs are fully used.