Problem
Reactive scaling causes outages. Over-provisioning wastes money. Need data-driven capacity decisions.
economics-evolutionestablished · medium operational burden
Capacity Planning
Also known as: capacity-model, resource-planning, scaling-model
Intent
Predict resource needs (CPU, memory, storage, network) for future load, enabling proactive scaling and cost optimization.
Forces
- •Load grows (users, data, features)
- •Resources have lead time (hardware, reserved instances)
- •Cost scales with capacity (right-size = save money)
- •Bursts: capacity for peak, not average
Solution
✓ When to Use
- •Growing systems (user base, data, features)
- •Cost optimization initiatives
- •Preparing for known events (Black Friday, launch)
- •Cloud commitment planning (RI, Savings Plans)
✗ When Not to Use
- •Stable, predictable workloads
- •Serverless (capacity managed by provider)
- •Team not ready for modeling discipline
Pros
- +Proactive: scale before saturation
- +Cost optimization: right-size, commit wisely
- +Risk reduction: known limits, planned headroom
- +Alignment: engineering ↔ finance ↔ product
Cons
- −Model uncertainty: growth assumptions wrong
- −Operational overhead: data collection, model maintenance
- −False precision: model looks exact, isn't
- −Reactive culture: planning seen as bureaucracy
Cost Profile
Infrastructure
Low — monitoring data, spreadsheet/model
Operational
Medium — monthly review, model updates
Cognitive
High — queueing theory, growth modeling
Failure Modes
Growth assumption wrong → capacity surprise
Model ignores new feature impact
Auto-scale too slow → outage before scale-out
Cost optimization → under-provisioned → outage
Single resource focus → bottleneck elsewhere
Real-World Examples
Alternatives
- auto-scaling-only
- reactive-scaling
- over-provisioning
- serverless
Related Patterns
- auto-scaling
- load-testing
- queueing-theory
- observability
- cost-optimization
Competency Domains
economics evolutionreliability opsdata statescalingdeployment