Core

MIG, MPS and Time-Slicing

A whole H100 is far more than a notebook, a small inference service or a CI job needs, and giving each of them a card leaves most of the fleet idle. Three mechanisms share a GPU, and they differ in what they isolate: MIG partitions the hardware into up to seven slices with their own memory and compute, MPS lets several processes share one GPU's SMs concurrently with no memory isolation, and time-slicing context-switches between processes with no isolation at all. The choice is the isolation the workload needs against the utilization the platform wants.

a free account unlocks the core curriculum tier · no card

RELATED CONCEPTS

Kubernetes GPU Scheduling Multi-Tenancy, Quotas and Fair Share GPU Execution Model GPU Memory Hierarchy Inference Autoscaling and Cold Starts Capacity Planning and Utilization

LESSONS THAT TEACH THIS

Fleet and ReliabilityNinety percent allocated and thirty percent used is the normal failure14 min→Foundations of AI InfrastructureA GPU pod fails differently from every other pod you have run14 min→Fleet and ReliabilityIsolation is what you sell, and the levels guarantee different things15 min→

PRACTICE THIS IN REAL QUESTIONS

Kubernetes, Slurm & GPU SchedulingMIG, time-slicing and MPS all let several jobs share one GPU. What is the difference, and when would you pick each?→GPU & Accelerator ArchitectureMIG versus MPS: what isolation does each give you when sharing a GPU, and which would you pick for a multi-tenant inference node?→Kubernetes, Slurm & GPU SchedulingDesign a notebook platform for 300 researchers on 64 GPUs. How do you share, reclaim and account for the GPUs?→Kubernetes, Slurm & GPU SchedulingDesign a GPU-aware scheduler that supports fractional GPUs: what isolation does each fraction get, and where does it break?→Kubernetes, Slurm & GPU SchedulingKubernetes device plugins versus Dynamic Resource Allocation: what changes for GPU scheduling, and what would you adopt in 2026?→Kubernetes, Slurm & GPU SchedulingDesign the scheduling and isolation for a multi-tenant fine-tuning service: hundreds of customers, a few base models, shared GPUs.→