AWSAmazon MWAAAnalytics

`aws_mwaa_environment` cost estimation

A managed Apache Airflow environment. Priced per environment-hour by size, plus additional worker and scheduler hours and metadata storage.

An aws_mwaa_environment runs Apache Airflow without you managing the scheduler, web server, or metadata database. It is priced on three things: the base environment, additional workers, and metadata storage. The base environment hour depends on the environment_class (mw1.small, mw1.medium, mw1.large), and it runs continuously, so even an idle environment bills 730 hours a month.

The base price includes one worker and one scheduler. Autoscaling workers beyond max_workers baseline, and extra schedulers, bill per additional unit-hour. For most teams the base environment dominates the bill: an mw1.small base is roughly $0.49/hour, about $358/month, before any extra workers.

c3x prices the base environment from environment_class. Additional worker hours and metadata storage are usage-based dimensions, since they depend on how much your DAGs actually scale; supply expected worker-hours in c3x-usage.yml for a fuller estimate. MWAA is one of those services where the floor cost surprises people, so seeing it before apply is the point.

Terraform example

A minimal but realistic configuration that C3X can estimate.

resource "aws_mwaa_environment" "data" {
  name              = "data-pipelines"
  environment_class = "mw1.small"

  airflow_version    = "2.10.1"
  dag_s3_path        = "dags/"
  execution_role_arn = aws_iam_role.mwaa.arn
  source_bucket_arn  = aws_s3_bucket.dags.arn

  max_workers = 10
  min_workers = 1

  network_configuration {
    security_group_ids = [aws_security_group.mwaa.id]
    subnet_ids         = aws_subnet.private[*].id
  }
}

Pricing dimensions

What you actually pay for when you provision aws_mwaa_environment.

Dimension	Unit	What's being charged
Environment hours	per hour	Base rate for the environment_class, billed continuously. Includes one worker and one scheduler. c3x assumes 730 hours/month. $0.49/hour for mw1.small
Additional worker hours	per worker-hour	Autoscaling workers beyond the included baseline. Usage-based; depends on DAG concurrency. Define expected worker-hours in c3x-usage.yml. $0.055/worker-hour for mw1.small
Additional scheduler hours	per scheduler-hour	Schedulers beyond the one included. Most environments run a single scheduler.
Metadata database storage	per GB-month	Airflow metadata DB storage. Usually small; usage-based. $0.10/GB-month

Sample C3X output

Example output from c3x estimate (base environment only):

aws_mwaa_environment.data
└─ Environment (mw1.small)              730  hours      $357.70

OVERALL TOTAL                                          $357.70
(additional workers usage-based; add to c3x-usage.yml)

Optimization tips

Common ways to reduce aws_mwaa_environment cost without changing the workload.

Right-size environment_class to scheduler load, not worker load

Up to 50% of base

The class sets scheduler/web-server capacity. If your DAGs are light but numerous, mw1.small with more autoscaling workers is cheaper than jumping to mw1.medium.

Tune min_workers to zero-idle where possible

Worker-hours during idle

MWAA scales workers down to min_workers when idle. Keeping min_workers at 1 avoids cold scheduling cost while letting the environment scale up for bursts.

Consider self-managed Airflow on ECS for very large fleets

Workload-dependent

Above a few large environments, running Airflow yourself on ECS/EKS can be cheaper, trading the managed convenience for control. Estimate both with c3x before deciding.

FAQ

How does c3x estimate MWAA cost?

It reads environment_class and prices the base environment hour at 730 hours/month. Additional workers, extra schedulers, and metadata storage are usage-based and added only when you provide expected values in c3x-usage.yml.

Why does an idle MWAA environment still cost ~$358/month?

The base environment (scheduler + web server + one worker) runs continuously and bills per hour whether or not DAGs are executing. There is no scale-to-zero for the base environment.

How are autoscaling workers priced?

Workers above the included baseline bill per worker-hour at a class-specific rate. Because actual worker-hours depend on DAG concurrency, c3x treats them as usage-based; set expected worker-hours in the usage file.

Does the Airflow version affect cost?

No. Pricing is by environment_class and worker/scheduler hours, not the Airflow version.

Is the underlying VPC / NAT cost included?

No. MWAA runs in your VPC, so any aws_nat_gateway or VPC endpoints it uses are billed by those resources. c3x estimates them separately if they are in your Terraform.