NVIDIA announced Workload Map, a common Kubernetes workload description that enables schedulers, controllers, and platforms to interpret distributed workloads without framework-specific parsing code.
What happened
NVIDIA introduced Workload Map, a common Kubernetes workload description that enables schedulers, controllers, and platforms to interpret distributed workloads without framework-specific parsing code. The project ships a catalog of prebuilt definitions for various frameworks, including JobSet, PyTorchJob, RayCluster, RayJob, RayService, KServe, Knative, MPIJob, NVIDIA NIM, LeaderWorkerSet, Milvus, Grove PodCliqueSet, and NVIDIA Dynamo. A single JobSet definition supports workload inspection through the Workload Map CLI, gang scheduling via KAI Scheduler, and unified management in NVIDIA Run:ai. Elotl's Nova uses Workload Map to interpret workload structure and resource requirements for multi-cluster scheduling decisions. The project is available under the Apache 2.0 license and is used in production within NVIDIA Run:ai.
What to weigh
- Workload Map supports workload inspection, scheduling, and platform visibility.
- Elotl’s Nova uses Workload Map for multi-cluster scheduling.
Source: NVIDIA