Unleash the power of resource scheduling for running Big Data & ML on Kubernetes!
![[object Object]](/img/resource-scheduling.png)
Builtin with comprehensive scheduling capabilities, including hierarchy queues, resource fairness across queues, job ordering (FIFO/FAIR), pluggable node sorting policies, preemption, and more.
![[object Object]](/img/support-k8s.png)
Fully K8s compatible, an alternative of the default K8s scheduler, but more powerful. Transparent for the existing K8s applications.
![[object Object]](/img/cloud-native.png)
Supports both on-prem and on-cloud use cases. When running on cloud, it works with the autoscaler to bring maximum resource elasticity, with better throughput.
Gives the fine-grained control over the resources quota for different tenants, easily map to your organization structure. The queue min/max capacity offers the over-commitment and ensures the guaranteed resources at the same time.
Recognize users, apps, queues. Apps are queued and scheduled with certain order. Apply the ordering based on resource fairness, submission time, and the priority. You get the more predictable scheduling results.
Optimized for Cloud, and accommodate the elasticity as much as possible. Gang Scheduling reduces the resource fragmentation, and triggers proactively up-scaling. Bin-packing turns the usage curve and lower your cost while running on Cloud.
No more lost tracking the resource usage of the tenants! YuniKorn provides a central management UI and a one-stop-dashboard to track the resource utilization of the cluster, queues, and apps. Start to plan and monitor the capacity for your teams!