Project Lightning Talk: From Static Slices To Elastic GPUs: Dynamic MIG With HAMI
Lightningat KubeCon + CloudNativeCon + OpenInfra Summit + PyTorch Conference China 2026 · 2026-09-07..09-09·
summary
Describes HAMI, a system that integrates the Kubernetes scheduler with the NVIDIA device plugin to change MIG partitions at scheduling time rather than pre-partitioning GPUs. Targets the problem of fragmentation and underutilization when operators must decide static MIG slice layouts for mixed, dynamic workloads. Shows an approach that adapts GPU partitioning dynamically within hardware constraints to improve flexibility and efficiency.
AI-written summary · about
related talks
10mLightning Talk: K8s Issue #52757: Sharing GPUs Among Multiple Containers
09m⚡ Lightning Talk: Advanced GPU-Orchestrated Workflows and HPC Integrations on K8s for Distributed AI/ML at Scale
04mLightning Talk: Dynamic GPU Fraction and Sharing With Cloud Native Principle
32mSmart GPU Management: Dynamic Pooling, Sharing, and Scheduling for AI Workloads in Kubernetes
30mGPU Sharing at CERN: Cutting the Cake Without Losing a Slice!
34mDeclarative Device Virtualization: Orchestrating GPUs & Hardware in Cloud Native Environments
05mHami: Dynamic, Smart, Stable GPU-Sharing Middleware In Kubern...
09mLightning Talk: Bringing Multi-Tenancy to GPU-Based Workloads