capacity-planning

Forecasting When Your Infrastructure Hits Capacity

Your service is growing 15% month-over-month but you only find out you need more nodes when latency spikes. This workflow collects 30 days of metrics, analyzes the growth trend, forecasts the day you run out of headroom, and recommends how many nodes to add and what it will cost.

Workflow

service, period
      |
      v
+--------------------+     +--------------------+     +--------------+     +----------------+
| cp_collect_metrics | --> | cp_analyze_trends  | --> | cp_forecast  | --> | cp_recommend   |
+--------------------+     +--------------------+     +--------------+     +----------------+
  43200 dataPoints          trend: "growing"          21 days until         action: "scale-up"
                            rate: 15%                 capacity              nodes: 3, $450/mo
                                                      confidence: 0.89

Workers

CollectMetricsWorker -- Takes service and period inputs. Returns dataPoints: 43200 representing 30 days of minute-level metrics.

AnalyzeTrendsWorker -- Receives the data point count. Determines the trend is "growing" at a rate of 15% month-over-month.

ForecastWorker -- Uses the trend to project that capacity will be exceeded in daysUntilCapacity: 21 with a confidence of 0.89.

RecommendWorker -- Based on the forecast, recommends action: "scale-up" by adding nodes: 3 at a cost of $450/month.

Tests

2 unit tests cover the end-to-end pipeline.

Running

See ../../RUNNING.md for setup and execution instructions.