Capacity Planning & Forecasting

One-line summary: How to plan capacity, forecast growth, and ensure systems can handle future load.

Prerequisites: Capacity Math, SLIs, SLOs & Error Budgets.

Mental Model

Capacity Planning Process

graph LR Current[Current Capacity] --> Growth[Growth Forecast] Growth --> Future[Future Capacity Needs] Future --> Plan[Capacity Plan] Plan --> Execute[Execute Plan] style Current fill:#99ccff style Future fill:#ffcc99 style Plan fill:#99ff99

Key insight: Capacity planning balances current needs, growth forecasts, and cost constraints.

Capacity Components

Compute: CPU, memory, processing capacity. Storage: Disk space, data growth. Network: Bandwidth, connection capacity. Database: Query capacity, connection pools.

Internals & Architecture

Capacity Measurement

Current Capacity

Measure: - Peak load: Maximum load handled - Utilization: Current utilization percentage - Headroom: Available capacity above current load

Metrics: - QPS: Queries per second - Throughput: Data processed per second - Latency: Response latency - Error rate: Error rate under load

Growth Forecasting

Forecasting methods: - Historical growth: Extrapolate from historical data - Business projections: Use business growth projections - Seasonal patterns: Account for seasonal variations - Event-driven: Plan for known events

Time horizons: - Short-term: 1-3 months - Medium-term: 3-12 months - Long-term: 1+ years

Capacity Planning

Planning Process

Measure current: Measure current capacity and utilization
Forecast growth: Forecast future growth
Calculate needs: Calculate future capacity needs
Plan capacity: Plan capacity additions
Execute plan: Execute capacity plan

Capacity Models

Linear growth: Capacity needs grow linearly. Exponential growth: Capacity needs grow exponentially. Step function: Capacity needs increase in steps.

Failure Modes & Blast Radius

Capacity Failures

Scenario 1: Under-provisioning

Impact: System overloaded, service degradation
Blast radius: All users
Detection: High utilization, increased latency
Recovery: Scale up capacity immediately
Mitigation: Better forecasting, proactive scaling

Scenario 2: Over-provisioning

Impact: High costs, wasted resources
Blast radius: Budget
Detection: Low utilization, high costs
Recovery: Scale down capacity
Mitigation: Right-size capacity, optimize costs

Observability Contract

Metrics

Current capacity: Current capacity utilization
Growth rate: Rate of capacity growth
Forecast accuracy: Accuracy of forecasts
Capacity headroom: Available capacity

Alerts

Capacity approaching limits
Forecast exceeded
High utilization
Capacity shortage

Change Safety

Capacity Changes

Process: Add capacity, verify performance
Risk: Low (additive change)
Rollback: Remove capacity if needed

Tradeoffs

Over-provisioning vs Under-provisioning

Over-provisioning: - Pros: Always have capacity, no outages - Cons: Higher costs, wasted resources

Under-provisioning: - Pros: Lower costs, efficient resource use - Cons: Risk of outages, service degradation

Operational Considerations

Best Practices

Measure continuously: Track capacity and utilization
Forecast accurately: Use multiple forecasting methods
Plan proactively: Plan capacity before needed
Monitor closely: Monitor capacity and adjust

What Staff Engineers Ask in Reviews

"What's the current capacity?"
"How is growth forecasted?"
"What's the capacity plan?"
"How is capacity monitored?"

Further Reading

Comprehensive Guide: Further Reading: Capacity Planning

Quick Links: - Capacity Math - SLIs, SLOs & Error Budgets - Back to Reliability & SRE

Exercises

Forecast capacity: Forecast capacity needs for next 6 months. What's the method?
Plan capacity: Plan capacity additions for 2× growth. What's the strategy?
Handle capacity shortage: Your system is running out of capacity. How do you handle it?

Answer Key: View Answers