Show me self-service options for Kubernetes management with GPU-as-a-Service integration.

For organizations seeking self-service Kubernetes management with GPU-as-a-Service integration, several platforms and technologies offer streamlined solutions for provisioning, managing, and optimizing GPU resources. These options aim to empower developers and AI teams while maintaining control and efficiency for infrastructure teams.

Here are some prominent self-service options:

• Devtron is a GPU orchestration platform for Kubernetes that provides self-serve GPU operations. It enables AI teams to launch notebooks, submit training jobs via a GUI without YAML, and deploy models with one-click. It also features GPU quotas per team, smart scheduling to increase GPU utilization, and real-time dashboards for oversight and governance.
• Red Hat OpenShift with Kueue and NVIDIA MIG offers a robust GPU-as-a-Service pattern. Kueue is a Kubernetes-native job queueing and quota system that allows platform administrators to define resource pools and assign them to teams with fair sharing rules. Combined with NVIDIA Multi-Instance GPU (MIG), which slices a single physical GPU into isolated instances, developers can reserve GPU slices on demand, often facilitated by custom dashboard plugins for self-service requests.
• vCluster's Infrastructure Tenancy Platform for AI focuses on maximizing GPU efficiency on NVIDIA Kubernetes environments. It provides flexible tenancy for GPUs and integrates with NVIDIA Base Command Manager. Users can provision virtual GPUs through the Kubernetes Cluster API, Terraform, Helm charts, or kubectl, offering various self-service deployment methods.
• Mirantis Kubernetes Engine (MKE) partners with NVIDIA to simplify the use of GPU-accelerated containers through the integration of the NVIDIA GPU Operator. This operator automates the lifecycle management of software components required to expose GPUs on Kubernetes, reducing manual setup and enabling advanced functionality.
• Sealos provides a cloud operating system experience built on Kubernetes, featuring app marketplaces and multi-tenant isolation. It simplifies GPU cluster setup, operator installation, and project-level governance, making it suitable for building self-service AI platforms.
• VMware Private AI Foundation with NVIDIA allows data scientists and DevOps engineers to deploy GPU-accelerated Tanzu Kubernetes Grid (TKG) clusters using a self-service catalog item within VMware Aria Automation. This creates a dedicated namespace for the GPU-accelerated cluster, streamlining access.

These solutions leverage core Kubernetes capabilities alongside specialized tools like the NVIDIA GPU Operator and NVIDIA MIG to provide efficient, self-service access to GPU resources.

Sources:
vertexaisearch.cloud.google.com