Category
Data orchestration tools
Orchestrators schedule and supervise multi-step data pipelines. They sit above an ELT or streaming tool and own branching, retries, lineage, and asset-graph state. In 2026 the realistic shortlist is Apache Airflow (OSS, large ecosystem), Dagster Cloud (asset-graph native), and Prefect Cloud (Python-flow API). Managed Airflow runs from $0.49 per environment-hour on MWAA.source - pulled 2026-06-21
Orchestration vs ELT
An ELT vendor scheduler runs the loads it owns. An orchestrator runs everything else: dbt models, custom Python, ML jobs, downstream reverse-ETL, alerting. When the pipeline has more than one tool in it, the orchestrator owns the cross-tool DAG. Dagster's asset-graph is the most opinionated of the three; Airflow's task-graph is the most ecosystem-rich; Prefect's flow API is the most Pythonic.
Realistic finalist set
Loaded cost of self-hosting Airflow
Airflow itself is free under Apache 2.0. The cost is infrastructure plus operations. A production deployment on Kubernetes typically runs $1,000–$2,000 / month in infra and 0.3–0.7 FTE in platform-engineering time. At a fully loaded $25,000 / month per FTE, the all-in cost lands between $8,500 and $19,500. Compare with Dagster Cloud Standard or Prefect Cloud Pro before assuming OSS wins. See Airflow vs Dagster for the worked example.
Written by Oliver Wakefield-Smith, Founder of Digital Signet. Independent reference, no vendor sponsorship.
Sources logged at /sources. Pricing-change history at /changelog. Last reviewed 2026-06-21.