Fractional GPUs, time-slicing and MIG — how the scheduler decides which pod gets silicon, and how to stop wasting it.
Kubernetes GPU Scheduling Explained
Fractional GPUs, time-slicing and MIG — how the scheduler decides which pod gets silicon, and how to stop wasting it.
Priya Nair9 min read
Published your local timeupdated
Written by
Priya Nair
Principal Editor, Cloud & Kubernetes
Priya writes about production Kubernetes and cloud-native architecture. Previously a staff SRE running multi-region clusters at scale.
More from Priya →