Skip to main content
Select from available boot source options to configure how your virtual machine will start and what operating system it will run.

​
Boot Source Types

Choose the boot source that best fits your deployment needs:

OS Images

Pre-configured operating systems optimized for GPU workloads
  • Ubuntu with CUDA pre-installed
  • Docker variants available
  • Optimized for AI/ML workloads
  • Ready to use immediately

Volumes

Boot from existing storage volume (authenticated users only)
  • Custom OS configurations
  • Persistent between deployments
  • Pre-installed software preserved
  • Minimum 100GB required

Custom Images

Use your own saved VM images (authenticated users only)
  • Created from your snapshots
  • Standardized environments
  • Team collaboration
  • Version-controlled deployments

​
OS Images

Pre-configured operating systems optimized for GPU workloads and machine learning applications.

​
Available OS Images

  • Ubuntu 22.04 LTS
  • Ubuntu 20.04 LTS
Latest Long-term Support ReleaseUbuntu Server 22.04 LTS R550 CUDA 12.4
  • NVIDIA Driver: R550 (latest stable)
  • CUDA Version: 12.4
  • Python: 3.10.x pre-installed
  • Package Manager: APT with GPU-optimized repositories
  • Best For: Latest AI frameworks, new projects
Ubuntu Server 22.04 LTS R550 CUDA 12.4 with Docker
  • All features above plus:
  • Docker: Latest Docker Engine pre-installed
  • Docker Compose: Ready for container workflows
  • NVIDIA Container Toolkit: GPU support in containers
  • Best For: Containerized AI/ML workloads

​
OS Image Features

All OS images include:
NVIDIA GPU Support
  • Latest NVIDIA drivers pre-installed
  • CUDA toolkit and libraries
  • cuDNN for deep learning acceleration
  • NCCL for multi-GPU communication
  • Optimized kernel parameters
Pre-installed Development Environment
  • Python 3 with pip and conda support
  • Essential build tools (gcc, make, cmake)
  • Git for version control
  • SSH server configured and running
  • Network tools and utilities
Machine Learning Ready
  • CUDA libraries for GPU computing
  • TensorRT for inference optimization
  • Essential system libraries
  • Optimized for popular ML frameworks
  • Ready for PyTorch, TensorFlow installation
Optimized System Settings
  • GPU-friendly kernel parameters
  • Optimized memory management
  • Network performance tuning
  • Security hardening applied
  • Automatic updates configured

​
Image Selection Guidelines

Choose Ubuntu 22.04

When You Need:
  • Latest software versions
  • Newest CUDA features
  • Modern Python 3.10+
  • Latest container support
  • New project development

Choose Ubuntu 20.04

When You Need:
  • Proven stability
  • Legacy software compatibility
  • Production environments
  • Extended support lifecycle
  • Conservative update approach

​
Bootable Volumes

Boot from an existing storage volume that contains a complete operating system and your custom configurations.

​
Volume Requirements

Bootable Volume Prerequisites:
  • Volume must be marked as “bootable” during creation
  • Minimum size: 100GB
  • Must contain a valid operating system
  • Volume and VM must be in the same region
  • Volume cannot be attached to another running VM

​
Creating Bootable Volumes

1

Create Volume

Navigate to Volumes and create a new volume:
  • Enable “Make Bootable” option
  • Select an OS image for initialization
  • Size must be at least 100GB
  • Choose the same region as your planned VM
2

Customize OS

Deploy a VM using the bootable volume:
  • Install required software and tools
  • Configure system settings
  • Set up applications and services
  • Create user accounts and permissions
3

Create Snapshot

Take a snapshot of the configured volume:
  • Ensures you have a backup of your configuration
  • Allows creating multiple identical volumes
  • Enables version control of your environment
4

Use for Deployment

Use the bootable volume for new VM deployments:
  • Select volume as boot source during deployment
  • Enjoy instant access to pre-configured environment
  • Consistent setup across multiple VMs

​
Bootable Volume Benefits

Custom Environment Preservation
  • All software installations persist
  • User accounts and permissions preserved
  • Application configurations maintained
  • System tweaks and optimizations retained
Fast VM Creation
  • Skip software installation time
  • Pre-configured environments ready immediately
  • Consistent setup across multiple deployments
  • Reduced time to productivity
Shared Development Environments
  • Identical environments for team members
  • Version-controlled system configurations
  • Standardized toolchains and libraries
  • Reduced environment-related issues
Environment Versioning
  • Multiple bootable volumes for different versions
  • Easy rollback to previous configurations
  • A/B testing of environment changes
  • Staged deployment strategies

​
Custom Images

Use your own saved VM images created from snapshots for standardized deployments.

​
Image Creation Process

1

Configure Source VM

Set up a VM with your desired configuration:
  • Install and configure all required software
  • Set up user accounts and permissions
  • Configure system settings and optimizations
  • Test all functionality thoroughly
2

Create Snapshot

Take a snapshot of the configured VM:
  • Navigate to Snapshots
  • Select your configured VM
  • Enable “Create as Image” option
  • Provide descriptive name and documentation
3

Verify Image

Test the custom image:
  • Deploy a new VM using the custom image
  • Verify all configurations are preserved
  • Test application functionality
  • Validate performance characteristics
4

Use for Production

Deploy production VMs using the custom image:
  • Consistent environments across deployments
  • Faster deployment times
  • Reduced configuration errors
  • Standardized production environments

​
Custom Image Advantages

Environment Consistency

Standardized Deployments
  • Identical software versions across VMs
  • Consistent configuration settings
  • Predictable performance characteristics
  • Reduced deployment variability

Operational Efficiency

Streamlined Operations
  • Faster VM deployment times
  • Reduced manual configuration
  • Lower risk of configuration errors
  • Simplified troubleshooting

​
Image Management Best Practices

Organized Image Management
  • Use descriptive names with versions
  • Include software versions in names
  • Document purpose and use cases
  • Follow team naming standards
Example naming:
  • ml-env-pytorch-2.1-cuda12.4-v1.0
  • web-app-prod-ubuntu22-v2.3
  • dev-env-team-alpha-v1.5
Image Lifecycle Management
  • Maintain multiple versions of images
  • Document changes between versions
  • Test new versions before production use
  • Plan deprecation strategy for old versions
Image Documentation
  • Document installed software and versions
  • Include configuration details
  • Provide usage instructions
  • Maintain change logs
Quality Assurance
  • Test images before production deployment
  • Validate functionality across different VM sizes
  • Test with different GPU types
  • Verify performance characteristics

​
Boot Source Selection Guide

​
Decision Matrix

  • Quick Start
  • Custom Environment
  • Standardized Deployment
OS ImagesChoose When:
  • Starting a new project
  • Need immediate GPU-ready environment
  • Want latest drivers and software
  • Prefer standard configurations
Best For: Development, experimentation, proof of concepts

​
Regional Considerations

Regional Availability:
  • OS images are automatically filtered by selected region
  • Volume and custom image options require user authentication
  • Bootable volumes must be in the same region as your VM
  • Custom images are created from your snapshots and are region-specific

​
Cross-Region Deployment

If you need to deploy across multiple regions:
1

Plan Multi-Region Strategy

  • Identify which boot sources you need in each region
  • Plan for image/volume replication needs
  • Consider regional compliance requirements
2

Create Regional Resources

  • Create bootable volumes in each target region
  • Generate custom images in required regions
  • Test deployment procedures in each region
3

Maintain Consistency

  • Keep images synchronized across regions
  • Document regional differences
  • Test cross-region compatibility

​
Performance Considerations

​
Boot Time Optimization

OS Images

Fastest Boot Times
  • Optimized for quick startup
  • Minimal additional configuration
  • Standard driver initialization
  • Ready for immediate use

Custom Sources

Configuration-Dependent
  • Boot time depends on customizations
  • Complex setups may increase boot time
  • Pre-installed software adds to startup
  • Consider startup script optimization

​
Storage Performance

Optimized Base Performance
  • Standard filesystem optimizations
  • GPU driver optimizations applied
  • Baseline performance characteristics
  • Predictable performance profile
Customizable Performance
  • Performance depends on volume configuration
  • Custom filesystem optimizations possible
  • May include performance-tuned applications
  • Consider volume size impact on performance
Inherited Performance Characteristics
  • Performance based on source VM configuration
  • May include custom optimizations
  • Consider image size impact on deployment
  • Test performance in target environment

​
Security Considerations

​
Boot Source Security

Security Best Practices:
  • OS images receive regular security updates
  • Custom images should be updated regularly
  • Bootable volumes require manual security maintenance
  • Verify source integrity before deployment

​
Security Validation

1

Verify Source Integrity

  • Use only trusted OS images
  • Validate custom image sources
  • Check bootable volume history
  • Verify no unauthorized modifications
2

Apply Security Updates

  • Update OS images regularly
  • Maintain custom images with security patches
  • Update bootable volumes periodically
  • Monitor security advisories
3

Configure Security

  • Apply appropriate security configurations
  • Configure firewalls and access controls
  • Set up monitoring and logging
  • Implement intrusion detection
Start with OS images for development and testing, then move to bootable volumes or custom images for production deployments that require specific configurations.
⌘I