VM States & Billing
Understanding virtual machine states and billing implications
Virtual machines can exist in various states, each with different billing implications. Understanding these states helps you manage costs effectively.
Primary VM States
ACTIVE
The VM is running with all resources allocated. This is the standard working state.
Billing: Full rate (per-minute billing)
SHUTOFF
The VM is stopped but resources remain reserved exclusively for your use.
Billing: Full rate (per-minute billing)
HIBERNATED
The VM’s state is saved while hardware resources are released. Only root disk storage remains allocated.
Billing: Reduced rate ($0.01/€0.01/₹1.00 per hour)
DELETED
The VM is permanently removed with all resources released.
Billing: No charge
Per-Minute Billing
The platform uses precise per-minute billing for optimal cost efficiency:
Billing Precision
Billing Precision
- Billing Calculation: Calculated in one-minute increments for exact usage tracking
- No Rounding: You pay only for the time your resources are actually running
- No Minimums: No minimum usage requirements or rounding to hours
- Real-time Tracking: Costs are tracked and displayed in real-time
State-Based Billing
State-Based Billing
- ACTIVE VMs: Billed at the full GPU rate per minute
- SHUTOFF VMs: Billed at the full GPU rate (resources remain reserved)
- HIBERNATED VMs: Billed at significantly reduced rate
- DELETED VMs: No billing charges
Cost Optimization
Cost Optimization
- Use Hibernation: For VMs not actively used but needed later
- Delete Unused VMs: For VMs no longer needed
- Monitor Usage: Track actual usage vs. allocated time
- Right-size Resources: Choose appropriate GPU types for workloads
Transitional VM States
During operations, your VM may temporarily be in one of these states. No additional charges apply during these transitional states:
Billing Rate Examples
GPU Billing Rates per Minute
Based on hourly rates, here are the per-minute costs:
| GPU Model | USD per Hour | USD per Minute |
|---|---|---|
| H100-SXM5-80GB | $2.69 | $0.0448 |
| H100-PCIe-NVLink-80GB | $2.29 | $0.0382 |
| H100-PCIe-80GB | $2.25 | $0.0375 |
| A100-SXM4-80GB-NVLink | $1.59 | $0.0265 |
| A100-PCIe-80GB | $1.55 | $0.0258 |
| L40 | $1.19 | $0.0198 |
| RTX-A6000 | $0.69 | $0.0115 |
| A40 | $0.69 | $0.0115 |
| GPU Model | USD per Hour | USD per Minute |
|---|---|---|
| H100-SXM5-80GB | $2.69 | $0.0448 |
| H100-PCIe-NVLink-80GB | $2.29 | $0.0382 |
| H100-PCIe-80GB | $2.25 | $0.0375 |
| A100-SXM4-80GB-NVLink | $1.59 | $0.0265 |
| A100-PCIe-80GB | $1.55 | $0.0258 |
| L40 | $1.19 | $0.0198 |
| RTX-A6000 | $0.69 | $0.0115 |
| A40 | $0.69 | $0.0115 |
| GPU Model | EUR per Hour | EUR per Minute |
|---|---|---|
| H100-SXM5-80GB | €2.49 | €0.0415 |
| H100-PCIe-NVLink-80GB | €2.09 | €0.0348 |
| H100-PCIe-80GB | €2.05 | €0.0342 |
| A100-SXM4-80GB-NVLink | €1.45 | €0.0242 |
| A100-PCIe-80GB | €1.39 | €0.0232 |
| L40 | €1.09 | €0.0182 |
| RTX-A6000 | €0.45 | €0.0075 |
| A40 | €0.45 | €0.0075 |
| GPU Model | INR per Hour | INR per Minute |
|---|---|---|
| H100-SXM5-80GB | ₹239 | ₹3.98 |
| H100-PCIe-NVLink-80GB | ₹199 | ₹3.32 |
| H100-PCIe-80GB | ₹195 | ₹3.25 |
| A100-SXM4-80GB-NVLink | ₹135 | ₹2.25 |
| A100-PCIe-80GB | ₹135 | ₹2.25 |
| L40 | ₹99 | ₹1.65 |
| RTX-A6000 | ₹49 | ₹0.82 |
| A40 | ₹49 | ₹0.82 |
Hibernation Benefits
Hibernation provides significant cost savings while preserving your VM state:
Cost Savings
Dramatic Cost Reduction
- Hibernated VMs cost only $0.01/€0.01/₹1.00 per hour
- Saves 95%+ compared to running VMs
- Perfect for VMs used intermittently
- Ideal for development environments
State Preservation
Complete Environment Preservation
- All applications and data remain intact
- Running processes are preserved
- Memory state is saved
- Quick restoration when needed
Hibernation Use Cases
Development Workflows
Development Workflows
Perfect for Development
- Preserve development environment between sessions
- Save costs during non-working hours
- Quick startup for continued work
- Maintain installed tools and configurations
Batch Processing
Batch Processing
Efficient for Scheduled Jobs
- Hibernate between batch job runs
- Preserve environment setup
- Cost-effective for periodic workloads
- Quick activation for scheduled tasks
Testing Environments
Testing Environments
Ideal for QA and Testing
- Preserve test environment states
- Cost-effective for intermittent testing
- Quick environment restoration
- Multiple test configurations available
Training Experiments
Training Experiments
Machine Learning Workflows
- Pause long-running training jobs
- Save costs during data preparation
- Preserve model checkpoints and environment
- Resume training when ready
Cost Management Strategies
Billing Optimization
For Brief Tasks (< 1 hour)
- Use per-minute billing advantage
- Choose appropriate GPU size
- Delete VM when complete
- Monitor usage in real-time
For Brief Tasks (< 1 hour)
- Use per-minute billing advantage
- Choose appropriate GPU size
- Delete VM when complete
- Monitor usage in real-time
For Development (daily use)
- Use hibernation during off-hours
- Right-size GPU for development needs
- Consider lower-cost GPUs for development
- Monitor and optimize usage patterns
For Production (continuous)
- Keep VMs in ACTIVE state for 24/7 workloads
- Use hibernation for scheduled maintenance
- Monitor performance vs. cost trade-offs
- Plan capacity based on usage patterns
For Extended Projects (weeks/months)
- Use hibernation for non-working hours
- Create snapshots for important milestones
- Consider bootable volumes for quick setup
- Regular cost monitoring and optimization
State Management Best Practices
Plan Your Usage
- Estimate actual usage hours per day
- Identify hibernation opportunities
- Plan for peak and off-peak usage
- Consider time zone differences for global teams
Implement State Management
- Use hibernation for regular off-hours
- Set up monitoring for usage patterns
- Automate hibernation where possible
- Create procedures for team members
Monitor and Optimize
- Track actual costs vs. estimates
- Identify optimization opportunities
- Adjust GPU sizes based on actual needs
- Regular review of hibernation effectiveness
Scale Efficiently
- Add VMs based on actual demand
- Use hibernation for overflow capacity
- Consider bootable volumes for quick scaling
- Monitor performance vs. cost metrics
Real-world Cost Examples
Example Scenarios
8-hour workday, 5 days per week
A100-PCIe-80GB:
- Active time: 40 hours/week × 62/week
- Hibernated: 128 hours/week × 1.28/week
- **Total: 260.40 if always active
- Savings: 76%
8-hour workday, 5 days per week
A100-PCIe-80GB:
- Active time: 40 hours/week × 62/week
- Hibernated: 128 hours/week × 1.28/week
- **Total: 260.40 if always active
- Savings: 76%
4 hours/day processing, 7 days per week
L40:
- Active time: 28 hours/week × 33.32/week
- Hibernated: 140 hours/week × 1.40/week
- **Total: 199.92 if always active
- Savings: 83%
12 hours/day intensive work, 7 days per week
H100-SXM5-80GB:
- Active time: 84 hours/week × 225.96/week
- Hibernated: 84 hours/week × 0.84/week
- **Total: 451.92 if always active
- Savings: 50%
24/7 continuous operation
A100-SXM4-80GB-NVLink:
- Always active: 168 hours/week × 267.12/week
- No hibernation benefits for 24/7 workloads
- Total: $267.12/week
- Optimization: Right-size GPU for actual load
Billing Monitoring
Real-time Cost Tracking
Dashboard Monitoring
Track Usage in Real-time
- Current VM states and costs
- Hourly and daily usage summaries
- Cost projections based on current usage
- Historical usage patterns
Cost Optimization Tools
Optimize Your Spending
- Hibernation opportunity identification
- Right-sizing recommendations
- Usage pattern analysis
- Cost-saving suggestions
Monitor your VM states and costs regularly through the dashboard to ensure optimal resource utilization and cost management.
Billing FAQ
How precise is per-minute billing?
How precise is per-minute billing?
Billing is calculated to the exact minute. If you use a VM for 37 minutes, you pay for exactly 37 minutes, not a full hour.
What happens during state transitions?
What happens during state transitions?
No additional charges apply during transitional states like STARTING, STOPPING, or HIBERNATING. You’re only billed when the VM reaches a stable state.
How quickly can I hibernate and restore?
How quickly can I hibernate and restore?
Hibernation typically takes 1-2 minutes, and restoration takes 2-3 minutes. The exact time depends on the VM’s memory usage and configuration.
Are there any hibernation limitations?
Are there any hibernation limitations?
Most VMs can be hibernated, but some workloads with specific hardware requirements may have limitations. The system will notify you if hibernation is not available.
How do I track my usage?
How do I track my usage?
Use the dashboard to monitor real-time usage, costs, and VM states. Historical usage data is also available for analysis.
Use hibernation strategically to reduce costs by up to 80% for workloads that don’t require 24/7 operation. This is especially effective for development and intermittent processing workloads.