Saturn Cloud on Crusoe

Deploy Saturn Cloud Enterprise on Crusoe’s AI cloud infrastructure

Saturn Cloud Enterprise for Crusoe provides a fully managed data science platform deployed directly into your Crusoe Cloud environment. Saturn Cloud Enterprise has a standard configuration for Crusoe Cloud. In a typical installation, we provision a Crusoe Managed Kubernetes (CMK) cluster in your Crusoe project, configure CPU and GPU node pools optimized for AI workloads, and integrate with Crusoe storage and networking services. Saturn Cloud is hosted publicly with secure authentication by default, with private networking options available if required.

This standard installation creates a ready-to-use platform for machine learning and AI workloads on sustainable, cost-effective GPU infrastructure. Beyond the default setup, Saturn Cloud can customize node pool sizes, GPU configurations, and network settings to meet your organization’s needs.

Architecture Overview

Saturn Cloud on Crusoe leverages Crusoe Managed Kubernetes (CMK) to provide a scalable, secure, and environmentally sustainable platform for data science teams.

The infrastructure includes:

Core Components

CMK Cluster: A fully managed Kubernetes cluster with automated control plane management
Sustainable Infrastructure: GPU compute powered by stranded energy and renewable sources
Node Pools: Auto-scaling node groups optimized for different workload types, including A100 GPU nodes
Storage: Integration with Crusoe CSI for persistent data storage
Container Registry: Seamless access to container images and registries

Network Architecture

Saturn Cloud integrates with Crusoe’s high-performance networking infrastructure:

1. Network Integration

Dedicated Networking: High-bandwidth, low-latency connections optimized for AI workloads
Location-Based Deployment: Deployed in your specified Crusoe location (us-east1-a, us-west1-a, etc.)
Managed Control Plane: Crusoe handles all networking for the Kubernetes control plane
SSH Access: Optional SSH access to nodes for debugging and management

2. Network Security

The Crusoe deployment ensures:

Secure API endpoints with TLS encryption
Network isolation between node pools and workloads
Kubernetes-native security policies
Optional VPC integration for enterprise networking requirements

Node Pool Architecture

The installation creates specialized node pools optimized for AI and data science workloads.

Note

The node pool configurations below are optimized for Crusoe’s AI infrastructure. Saturn Cloud can customize these configurations to match your specific requirements, including:

Different CPU and GPU instance types
Custom memory and storage configurations
Specific location preferences
Tailored autoscaling policies
Mixed workload optimizations

Contact support@saturncloud.io to discuss customization options for your deployment.

1. System Node Pool

Purpose: Runs Saturn Cloud control plane and system services
Instance Type: c1a.4x (16 vCPUs, 32GB RAM)
Scaling: Fixed 1-2 nodes for system stability
Storage: High-performance local storage

2. CPU Node Pools

Saturn Cloud provisions CPU-optimized node pools for various computational requirements:

Compute-Optimized (c1a.4x)

Configuration: 16 vCPUs, 32GB RAM
Use Case: General purpose workloads, data preprocessing, development environments
Auto-scaling: 0-100 nodes based on demand
Cost: Optimized pricing for sustained workloads

3. GPU Node Pools

Saturn Cloud provides access to Crusoe’s full GPU fleet, including NVIDIA A100, H100, and L40 GPUs. Our standard configuration showcases A100 80GB configurations, though Saturn Cloud can provision any GPU type available in Crusoe’s infrastructure to meet your specific requirements.

Single A100 80GB (a100-80gb.1x)

Configuration: 1x A100 80GB GPU, 16 vCPUs, 120GB RAM
GPU Memory: 80GB HBM2e per GPU
Use Case: Model development, inference, single-GPU training
Auto-scaling: 0-100 nodes
Sustainability: Powered by stranded energy sources

Dual A100 80GB (a100-80gb.2x)

Configuration: 2x A100 80GB GPUs, 32 vCPUs, 240GB RAM
GPU Memory: 160GB total HBM2e
Use Case: Multi-GPU training, larger model development
Auto-scaling: 0-50 nodes

Quad A100 80GB (a100-80gb.4x)

Configuration: 4x A100 80GB GPUs, 64 vCPUs, 480GB RAM
GPU Memory: 320GB total HBM2e
Use Case: Large-scale model training, distributed computing
Auto-scaling: 0-25 nodes

Octa A100 80GB (a100-80gb.8x)

Configuration: 8x A100 80GB GPUs, 128 vCPUs, 960GB RAM
GPU Memory: 640GB total HBM2e
Use Case: Enterprise-scale training, LLM development, research workloads
Auto-scaling: 0-10 nodes

GPU Fleet Access

The A100 configurations shown above represent our reference implementation. Crusoe offers additional GPU types including H100 and L40 GPUs. Contact support@saturncloud.io to discuss accessing Crusoe’s complete GPU portfolio for your specific workload requirements.

Sustainable AI Computing

Crusoe’s GPU infrastructure delivers exceptional sustainability benefits:

90% lower carbon footprint compared to traditional cloud providers
Stranded energy utilization - computing powered by otherwise wasted energy
Renewable energy integration - supporting clean energy grid stability
Cost efficiency - up to 50% lower costs than hyperscaler GPU offerings

Security Features

Saturn Cloud on Crusoe is deployed with multiple security controls to protect access, data, and workloads:

Managed Authentication: Secure API access through Crusoe’s authentication system
SSH Key Management: Optional SSH access with your public keys
Network Isolation: Kubernetes-native network policies and security groups
Encrypted Storage: All persistent storage encrypted at rest
Secure Endpoints: TLS-encrypted communication for all services
RBAC Integration: Role-based access control for Kubernetes resources

Storage and Data Access

Saturn Cloud leverages Crusoe storage to ensure reliable data access and high availability:

Crusoe CSI: High-performance Container Storage Interface for persistent volumes
Local Storage: Fast local NVMe storage for temporary workloads
External Integration: Support for external storage services and data sources
Snapshot Support: Volume snapshots for backup and recovery

Monitoring and Operations

Operational features in Saturn Cloud automate scaling, resource allocation, and system health management:

Cluster Autoscaler: Automatic node provisioning based on workload demands
GPU Operator: NVIDIA GPU resource management and scheduling
Network Operator: Optimized networking for GPU workloads
Health Monitoring: Continuous monitoring of node and cluster health
Performance Insights: Built-in monitoring and performance analytics

Getting Started

Ready to revolutionize your AI infrastructure with sustainable computing? contact our team to design your custom Crusoe deployment.

Next Steps:

Sustainability Assessment: Evaluate your current carbon footprint and cost savings potential
Architecture Design: Create an optimized deployment for your AI workloads
Pilot Deployment: Start with a proof-of-concept in your Crusoe environment
Production Scale: Expand to full production with ongoing optimization support

Limited Sustainable Capacity

Sustainable GPU capacity is in high demand as organizations prioritize ESG compliance. Contact support@saturncloud.io today to secure your allocation and begin your transition to carbon-neutral AI computing.

Installation Process

Saturn Cloud Enterprise for Crusoe is fully managed. The installation steps are:

Prerequisites

Crusoe Cloud Account: Sign up at https://console.crusoecloud.com/
API Credentials: Create API keys from the Crusoe console security section
Location Selection: Choose your preferred Crusoe location for deployment

Installation Steps

Prepare Crusoe Credentials: Set up API credentials in your environment:

# Add to ~/.crusoe/config
[default]
access_key_id="YOUR_ACCESS_KEY"
secret_key="YOUR_SECRET_KEY"

Configure Installation Parameters: Provide the following information to Saturn Cloud support:

# Crusoe project configuration
project_id = "xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx"
cluster_name = "saturn-cluster-crusoe"
location = "us-east1-a"
kubernetes_version = "1.30.8-cmk.28"

Grant Installation Access: Provide Saturn Cloud support with necessary API permissions
Infrastructure Deployment: Our team handles the CMK cluster deployment and configuration
Access Your Platform: Receive your Saturn Cloud URL and admin credentials

Ongoing Management

Saturn Cloud Enterprise on Crusoe is fully managed, meaning:

Automatic Kubernetes and GPU driver updates
24/7 monitoring and support for sustainable AI workloads
Performance optimization and cost monitoring
Sustainability reporting and carbon footprint tracking
Backup and disaster recovery
Continuous optimization for cost and environmental impact

Regional Availability

Saturn Cloud on Crusoe can be deployed in locations where Crusoe Cloud operates:

us-east1-a (Virginia) - Primary region with full A100 availability
us-west1-a (California) - West coast deployment option
Additional locations as Crusoe expands sustainable infrastructure

Cost and Sustainability Optimization

The Crusoe infrastructure provides unique optimization advantages:

Auto-scaling to zero - minimize costs and environmental impact when idle
Stranded energy utilization - lowest possible carbon footprint
Efficient resource allocation - right-sized instances for different workloads
Sustainability reporting - detailed carbon footprint tracking
Cost transparency - clear, predictable pricing with no hidden fees
Green computing incentives - potential tax benefits for sustainable computing

Performance and Environmental Benefits

Saturn Cloud on Crusoe delivers unique advantages:

Carbon-neutral AI computing - industry-leading sustainability
Cost-effective GPU access - premium A100, H100, and L40 GPUs at accessible prices
High-performance networking - optimized for AI workloads
Renewable energy integration - supporting clean energy infrastructure
Enterprise reliability - 99.9% uptime with sustainable operations