Work out etl job cost instantly with clear inputs, formula shown and shareable results.
ETL cost is worker-hours multiplied by the worker rate, and because duration falls as workers rise, the cost per run is roughly constant with parallelism — you buy speed, not savings. The real levers are processing less data through incremental loads and partition pruning, and cost per terabyte processed is the metric to track over time.
ETL cost
run hours = data / (workers x per-worker throughput); cost per run = run hours x workers x rate; monthly = cost per run x runs x 30.44
Not directly, since cost is worker-hours. It reduces wall-clock time, and can reduce cost slightly by cutting fixed startup overhead as a share of the run.
Incremental processing. Reprocessing the full dataset every run is the single most common and most expensive ETL antipattern.