dstack Free plan 8.9 Visit 
HAMi
project-hami.io
What it is
HAMi is open-source middleware for sharing and scheduling GPUs and other accelerators among AI workloads on Kubernetes. Workloads can request GPU memory and core limits through Kubernetes resource controls. Scheduling options include binpack, spread and numa-first, with binpack and spread available at node or GPU scope. The supported-device list spans NVIDIA, AMD, AWS Neuron, Huawei Ascend, Cambricon, Enflame, Hygon, Iluvatar, Kunlunxin, MetaX, Moore Threads, Vastai and Biren; isolation and partitioning capabilities vary by device. HAMi works with Kubernetes APIs, DRA and CDI, and documents scheduling integrations including Volcano, Kueue, Koordinator and NVIDIA KAI Scheduler. Its node metrics endpoint exposes device memory and core utilization for Prometheus. Classic HAMi requires Kubernetes v1.23 or later. HAMi-DRA requires Kubernetes v1.34 or later, the DRA Consumable Capacity feature gate, CDI and NVIDIA driver 440 or later. NVIDIA limits use user-space library interception, which does not cover applications that bypass CUDA. Enforcement is described as soft and best-effort; the FAQ recommends MIG when hardware-enforced isolation is required.
Spec sheet
- Purpose
- HAMi is open-source, cloud-native GPU virtualization middleware for sharing, isolating, and scheduling heterogeneous accelerators on Kubernetes.Read 4 Oct 2026 · project-hami.io
- Product
- HAMi is open-source GPU virtualization middleware that enables sharing, isolation and scheduling of heterogeneous accelerators for AI workloads on Kubernetes.Read 3 Oct 2026 · project-hami.io
- Deployment
- The quick start installs HAMi with Helm and requires Kubernetes, Helm, kubectl, and installation permissions.Read 4 Oct 2026 · project-hami.io
- Integrations
- The site lists Kubernetes, Volcano, Kueue, Koordinator and KAI Scheduler in its Kubernetes scheduling ecosystem.Read 3 Oct 2026 · project-hami.io
- Support
- The site links to documentation, tutorials, Discord and the `#hami-dev` Slack channel for community resources.Read 3 Oct 2026 · project-hami.io
- Accelerators
- The project lists support for NVIDIA, AWS Neuron, Huawei Ascend, Cambricon, Enflame, Hygon, Iluvatar, Kunlunxin, MetaX, Moore Threads, Vastai, AMD and Biren accelerators.Read 3 Oct 2026 · project-hami.io
- Community support
- The project links to community support through Discord and Slack (#hami-dev).Read 4 Oct 2026 · project-hami.io
- Deployment requirements
- Classic HAMi supports Kubernetes v1.23 or later; HAMi-DRA requires Kubernetes v1.34 or later with the DRA Consumable Capacity feature gate enabled, CDI and NVIDIA driver 440 or later.Read 3 Oct 2026 · project-hami.io
- Device coverage
- The v2.10.0 supported-device matrix lists NVIDIA, Cambricon, Hygon, Huawei Ascend, Iluvatar, Mthreads, MetaX, Enflame, Kunlunxin, Vastai, AMD, AWS Neuron, and Biren devices as stable.Read 4 Oct 2026 · project-hami.io
- Ecosystem integrations
- The project documents integrations with Volcano, Kueue, Koordinator, and NVIDIA KAI Scheduler.Read 4 Oct 2026 · project-hami.io
- Isolation limits
- The FAQ characterizes HAMi vGPU memory and compute enforcement as soft and best-effort, and recommends MIG when hardware-enforced isolation is required for compliance or SLAs.Read 3 Oct 2026 · project-hami.io
- Isolation mechanism
- For NVIDIA devices, HAMi enforces limits through user-space library interception; the FAQ says applications that bypass the CUDA library are not covered.Read 3 Oct 2026 · project-hami.io
- Kubernetes integration
- HAMi works with Kubernetes APIs, DRA, and CDI.Read 4 Oct 2026 · project-hami.io
- Limitations
- The supported-device matrix marks memory isolation, core isolation, and multi-card partitioning as unavailable for some listed devices.Read 4 Oct 2026 · project-hami.io
- Monitoring
- HAMi provides allocation counts and spread plus real-time GPU memory and core utilization visibility.Read 4 Oct 2026 · project-hami.io
- Project status
- HAMi is a CNCF Incubating project.Read 4 Oct 2026 · project-hami.io
- Resource controls
- HAMi supports GPU memory and compute quotas, with hard runtime isolation for supported devices.Read 4 Oct 2026 · project-hami.io
- Resource slicing
- HAMi lets workloads request GPU memory and core limits using Kubernetes resource limits such as `nvidia.com/gpumem` and `nvidia.com/gpucores`.Read 3 Oct 2026 · project-hami.io
- Scheduling
- HAMi offers binpack, spread, and topology-aware scheduling policies.Read 4 Oct 2026 · project-hami.io
- Scheduling limit
- HAMi's built-in priority field supports two levels; the FAQ recommends integrating Volcano for multi-level queue priorities.Read 3 Oct 2026 · project-hami.io
- Workloads
- The site identifies LLM, machine-learning, and HPC workloads as use cases.Read 4 Oct 2026 · project-hami.io
Plans and prices
HAMi Free Open-source GPU virtualization middleware for AI workloads on Kubernetes Read 4 Oct 2026 · project-hami.io
Strong points and limits
Strong points
- Free, open-source middleware.
- Supports a broad list of accelerator vendors.
- Offers binpack, spread and numa-first scheduling.
- Exposes utilization metrics for Prometheus scraping.
Limits
- Classic deployment requires Kubernetes v1.23 or later.
- NVIDIA interception does not cover apps bypassing CUDA.
- Memory and compute enforcement is soft and best-effort.
- Some devices lack isolation or partitioning capabilities.
Who it suits
HAMi suits teams running AI, machine-learning or HPC workloads on Kubernetes that need to share accelerator resources. It is free and self-hosted, but deployment requirements and device-specific isolation limits matter.
Verdict
HAMi provides Kubernetes-based accelerator sharing, quotas and scheduling across a wide device list. Check the requirements and per-device capability matrix, and consider MIG where hardware-enforced isolation is necessary.
Where Specifiction places it
- GPU Cluster Management Software#5 of 218.3
