HAMi vs OpenPBS
HAMi
project-hami.ioOpenPBS
openpbs.orgSide by side
| Figure | HAMi | OpenPBS |
|---|---|---|
| Price from | Free | Free |
| Free plan | Yes | Yes |
| Free trial | — | — |
| Open source | — | — |
| macOS | — | Yes |
| Linux | Yes | Yes |
| API | — | Yes |
| Self-hosted | Yes | Yes |
| Plans published | 1 | 1 |
| Spec score · GPU Cluster Management Software | 8.3 (#5) | 7.6 (#9) |
| Community support | The project links to community support through Discord and Slack (#hami-dev). | Community support is provided through a volunteer forum with no fee and no guarantee that questions will be answered or bugs addressed. |
| Purpose | HAMi is open-source, cloud-native GPU virtualization middleware for sharing, isolating, and scheduling heterogeneous accelerators on Kubernetes. | OpenPBS optimizes job scheduling and workload management in high-performance computing environments. |
| Scheduling | HAMi offers binpack, spread, and topology-aware scheduling policies. | Policy-driven scheduling balances job turnaround time and utilization with optimal job placement. |
Plans
In their own words
HAMi
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.
OpenPBS
OpenPBS schedules jobs and manages workloads in high-performance computing environments, including clusters, clouds, and supercomputers. Its policy-driven scheduling balances job turnaround time and system utilization while selecting job placements. The system is designed for millions of cores and has been tested beyond 50,000 nodes. A plugin framework provides administration visibility and supports customized extensions. Health checks monitor faults and can mitigate them automatically; the fail-over architecture is designed to avoid a single point of failure and keep jobs running through failures. OpenPBS is self-hosted and free under the AGPL 3.0 open-source license, while PBS Pro is also dual licensed commercially. Public-sector organizations such as academic and research facilities generally choose the open-source license. Community help comes through a volunteer forum, without a guarantee of answers or bug fixes. OpenPBS is integrated into the OpenHPC software stack. The installation guide states that SELinux is not supported, and critical and high security fixes are delivered in the next major release.