Side by side

Apache YuniKorn vs HAMi

Both stand on gpu cluster management software. Which one you can start free with, which one lets you read the code, and what each costs once you pay.

Apache YuniKorn

Free plan APISelf-hostedWeb

YuniKorn offers Kubernetes workload scheduling with tenant queues and several scheduling policies. Teams should account for deployment on an existing cluster and the build step for source releases.

Visit Apache YuniKorn

HAMi offers accelerator sharing, quotas, scheduling policies, and utilization metrics for Kubernetes workloads. Check device-specific capabilities and isolation limits, especially where hardware-enforced isolation is required.

Visit HAMi

The facts

	Apache YuniKorn	HAMi
Free plan	Yes	Yes
Open source	Not published	Not published
Runs on	API, Self-hosted, Web	Linux, Self-hosted
Lowest paid plan	Free or on request	Free or on request
Plans published	1	1
Maker	Not published	Not published
Website	yunikorn.apache.org	project-hami.io
On gpu cluster management software	#2 of 21	#6 of 21

Apache YuniKorn: what it does well

Free under the Apache License 2.0.
Hierarchical queues provide tenant resource controls.
Supports gang scheduling, reservations, and preemption.
Web UI shows capacity, utilization, and applications.
Metrics can be exposed via Prometheus.

What to know first

Source releases must be built before use.
Deployment guidance assumes an existing Kubernetes cluster.
No free trial is listed.

HAMi: what it does well

GPU memory and core quotas for workloads
Scheduling includes binpack, spread, and numa-first
Metrics endpoint can be scraped by Prometheus
Integrates with Kubernetes APIs, DRA, and CDI

What to know first

Classic deployment requires Kubernetes v1.23 or later
HAMi-DRA has Kubernetes v1.34 and additional requirements
NVIDIA enforcement does not cover applications bypassing CUDA
Memory and compute enforcement is soft and best-effort

Alternatives to Apache YuniKorn Alternatives to HAMi The whole list