Scheduling and control plane system focused on GPU placement, fragmentation, and topology-aware workload orchestration across heterogeneous fleets.
- Multi-tenant, intent-driven GPU scheduling for Kubernetes
- Prevents GPU fragmentation and preserves topology for large jobs
- Integrates with Kueue for admission and quota management
- Supports NUMA, network fabric, and workload-aware placement
- kubenexus-scheduler: Topology- and fragmentation-aware scheduler for multi-tenant GPU fleets.
- kubenexus-control-plane: Workload intent API, placement planning, and explainability for GPU-accelerated Kubernetes clusters.
See our repositories for schedulers, control plane, and deployment tools.
Apache 2.0 License.