Saturn Cloud
UnknownWhite-labeled control plane for GPU cloud operators
工具信息
什么是 Saturn Cloud？
Saturn Cloud is a white-labeled control plane for GPU cloud operators. It helps operators provision, operate, meter, and bill multi-tenant GPU infrastructure under their own brand.
核心功能
Multi-tenant isolation
Provides per-tenant Kubernetes or Slurm clusters, storage namespaces, SSO federation, RBAC, IAM, audit logging, and hardware-boundary isolation.
GPU operations support
Includes 24/7 on-call support, workload-aware rolling upgrades, GPU-specific monitoring, automatic node cordoning, contractual SLOs, and incident escalation.
Metering and billing
Tracks GPU-hour and token usage by user, project, and tenant, applies contract terms, and provides invoicing data through an API and Snowflake view.
Workload orchestration
Orchestrates bare metal, managed Kubernetes, managed Slurm, GPU development workspaces, training and fine-tuning jobs, inference workloads, and token factory endpoints.
GPU provisioning
Lets customers choose GPU hardware, configure cluster size, and launch workloads through the console.
Developer workspaces
Supports GPU development workspaces using JupyterLab, VS Code, SSH, and persistent storage.
使用场景
- #1
Operating a branded GPU cloud
GPU cloud operators can offer infrastructure under their own brand using Saturn Cloud as the control plane.
- #2
Multi-tenant GPU infrastructure
Operators can provision isolated Kubernetes or Slurm environments and separate storage, networking, identity, and hardware resources for each customer.
- #3
Managed GPU operations
Organizations can delegate GPU infrastructure incident response, monitoring, upgrades, and vendor escalation to Saturn Cloud.
- #4
GPU usage chargeback and invoicing
Finance and operations teams can meter usage by user, project, and tenant, apply contract terms, and export invoicing data for reporting.
- #5
Model training and inference
Customers can use GPU infrastructure for interactive development, fine-tuning, distributed training, managed jobs, and inference endpoints.
定价方案
The source snapshot links to Plans & Pricing and shows an example GPU rate of $2.95 per hour for listed hardware, but it does not provide complete plan or pricing terms.
