CloudWatch alarms pricing: what to model (alarm-month by type)

CloudWatch alarms pricing is straightforward once you model the right unit: alarm-month. Most estimation mistakes come from mixing alarm types or ignoring short-lived environments that still generate billable alarm-months.

Alarm pricing inputs

• Standard alarms: count per metric and dimension.
• High-res: 1-minute or 10-second alarms.
• Composite: add-on charges by count.

What to include in an estimate

• Standard alarm-months
• High-resolution alarm-months (if used)
• Composite alarm-months (if used)
• Downstream notifications: SNS topics, SMS, email, chat webhooks (separate services)

A fast modeling workflow

1. Count current alarms by type (standard/high-res/composite).
2. Convert ephemeral environments into time-weighted alarm-months.
3. Apply your region pricing to each alarm type.
4. Add notification delivery costs separately (if meaningful for your volume).

How to separate alarm types (so rates are applied correctly)

• Standard vs high-resolution: treat these as different buckets if your pricing differentiates.
• Composite alarms: count them separately; composites are useful to reduce paging noise but still add alarm-months.
• Non-prod: model prod and non-prod separately; many teams can reduce non-prod alarm packs safely.

Worked example (alarm-month math)

• Production: 300 standard alarms active all month -> 300 alarm-months
• PR envs: 40 alarms per env, ~10 envs active per day -> 400 alarms/day
• Monthly PR alarm-months ~= 400 * 30.4 / 30.4 ~= 400 alarm-months

This is why “small” PR environments can dominate alarm-months if they create the full production alarm pack.

Don't forget adjacent costs

• SNS delivery: if you send SMS/email or high-volume notifications, model it separately.
• Metrics and dashboards: alarms are often created alongside extra metrics and dashboards that add costs too.

Common pitfalls

• Per-instance alarms at scale: alarm count grows with fleet size.
• Full alarm packs in PR environments: short-lived stacks still add alarm-months.
• Duplicate monitoring stacks: two tools create equivalent alarms for the same incident.
• Ignoring notification plumbing: SMS and high-volume messaging can be a separate cost driver.

Optimization direction

If cost is high, do not start by deleting critical alarms. Start by removing unused alarms and replacing per-resource alarms with service-level signals.

Guide: CloudWatch alarms cost optimization.

Validation checklist

• Validate alarm counts from inventory and from IaC (two sources if possible).
• Separate prod vs non-prod and pro-rate ephemeral environments.
• Confirm alarm type usage (standard vs high-res vs composite) before applying rates.

Sources

• CloudWatch pricing: aws.amazon.com/cloudwatch/pricing

Related guides

Related calculators

FAQ