How to Cut AWS Config Recording Costs, Step by Step

Published by Rushikesh Dhule, LessBill · November 22, 2026 · Prices checked on this date
AWS Config is one of those services people switch on once, for compliance, and then forget. The bill does not forget: it charges for every configuration change it records, so an account full of short-lived pods, network interfaces or autoscaled instances can quietly pay hundreds of dollars a month just to keep a history nobody reads.
The fix is usually not to turn Config off. It is to find the two or three resource types that create most of the records, then either record them once a day or stop recording them. This guide walks through that in five steps, with the console paths and CLI commands, and the maths for deciding which option is cheaper.
What you are actually paying for
According to the AWS Config pricing page (US East, N. Virginia, checked November 22, 2026), there are three main charges:
- Configuration items (CIs). A CI is a snapshot of one resource's configuration. With continuous recording you pay $0.003 per CI, one for every change. With periodic (daily) recording you pay $0.012 per CI, but you get at most one per resource per day.
- Config rule evaluations. $0.001 each for the first 100,000 evaluations a month.
- Conformance pack evaluations. Also $0.001 each for the first 100,000 a month.
On top of that you pay normal S3 rates for the snapshot and history files Config delivers, SNS for change notifications, and Lambda for any custom rules.
Daily recording arrived in November 2023. The launch announcement explains that a daily CI captures the latest state of a resource over the past 24 hours, and only if it differs from the last one recorded. You can set a default frequency and then override it per resource type, so one account can record IAM roles continuously and network interfaces daily.
Step 1: Confirm Config is a line worth chasing
In the Billing console, open Cost Explorer, filter Service to Config, and group by Usage type. You will see separate lines for configuration items, rule evaluations and conformance pack evaluations, per region. If Config is a few dollars a month, stop here. If it is in the hundreds, or it jumped after a new cluster or deployment pipeline went live, keep going. A sudden jump is exactly the kind of thing daily cost-spike alerts should catch.
Step 2: See how your recorder is set up today
Run this in each region where Config is enabled:
aws configservice describe-configuration-recorders
Look at two parts of the output. recordingGroup tells you whether you record all supported types ("allSupported": true) or a chosen list. recordingMode shows the default recordingFrequency (CONTINUOUS or DAILY) and any per-type overrides. Most accounts set up years ago show all types, continuous, no overrides.
Step 3: Find the noisy resource types
AWS Config publishes usage metrics to CloudWatch. Since April 2024, those metrics show only billable usage, so they line up with your bill (AWS announcement). In the console: CloudWatch → Metrics → All metrics → Config, then the resource type dimension, and sort by the sum over the last 30 days.
From the CLI, first list what is available, then pull a 30-day daily sum for any type you suspect:
aws cloudwatch list-metrics --namespace AWS/Config
aws cloudwatch get-metric-statistics \
--namespace AWS/Config \
--metric-name ConfigurationItemsRecorded \
--dimensions Name=ResourceType,Value=AWS::EC2::NetworkInterface \
--start-time 2026-10-23T00:00:00Z \
--end-time 2026-11-22T00:00:00Z \
--period 86400 \
--statistics Sum
The usual suspects are types that are created, changed and destroyed by automation: AWS::EC2::NetworkInterface (every pod IP on EKS with the VPC CNI, every Lambda in a VPC scaling up), AWS::EC2::Instance in busy Auto Scaling groups, security groups edited by pipelines, and tags rewritten on every deploy. If your Kubernetes nodes churn a lot, expect network interfaces near the top.
Step 4: Do the maths before you change anything
Daily CIs cost four times more each ($0.012 ÷ $0.003 = 4). So daily recording only saves money for a resource that changes more than four times on the days it changes. A resource that changes once a day costs $0.003 continuous but $0.012 daily: switching it would quadruple the cost.
Here is a worked example. The numbers are an illustration, not a real customer: an account in us-east-1 records 600,000 CIs a month, of which 380,000 are network interfaces. About 2,000 interfaces exist on a typical day.
- Total CI cost today: 600,000 × $0.003 = $1,800/month.
- Network interfaces alone: 380,000 × $0.003 = $1,140/month.
- Changes per interface per day: 380,000 ÷ 30 ÷ 2,000 ≈ 6.3. That is above 4, so daily recording wins.
- Daily recording, worst case (every interface recorded every day): 2,000 × 30 × $0.012 = $720/month. Real usage will be lower, because unchanged resources produce no CI.
- Excluding the type entirely: $0 for CIs. Since the April 2024 change, Config still tracks creation and deletion of non-recorded resources at no charge.
| Option for network interfaces | Monthly CI cost | What you give up |
|---|---|---|
| Continuous (today) | $1,140 | Nothing: every change recorded |
| Daily override | up to $720 | Intra-day history; change-triggered rules run less often |
| Excluded | $0 | All configuration history and rule checks for that type |
In this example the total bill drops from $1,800 to at most $1,380 with a daily override (a 23% cut), or to $660 if you exclude the type.
Step 5: Apply the change
Console. The AWS Config documentation gives the path: open the AWS Config console, choose Settings, open the Customer managed recorder tab, edit it, choose All resource types with customizable overrides, pick the default frequency, add an override for the noisy type, and save.
CLI. Put the recorder settings in a file. This keeps everything continuous except network interfaces (replace the role ARN with your own from Step 2):
cat > recorder.json <<'EOF'
{
"name": "default",
"roleARN": "arn:aws:iam::123456789012:role/config-role",
"recordingMode": {
"recordingFrequency": "CONTINUOUS",
"recordingModeOverrides": [
{
"description": "ENIs churn with pods, record daily",
"recordingFrequency": "DAILY",
"resourceTypes": ["AWS::EC2::NetworkInterface"]
}
]
}
}
EOF
aws configservice put-configuration-recorder \
--configuration-recorder file://recorder.json \
--recording-group file://group.json
For group.json, copy the recordingGroup block from Step 2 so you do not change what is recorded by accident. To exclude a type instead, the CLI guide shows a group with "allSupported": false, an exclusionByResourceTypes list and "recordingStrategy": {"useOnly": "EXCLUSION_BY_RESOURCE_TYPES"}. Run describe-configuration-recorders again afterwards to check, then watch the CloudWatch metric for a week.
When not to do this
- Audit needs every change. If an auditor or your incident process relies on a full change history for security groups, IAM or KMS keys, keep those continuous. Save money on the noisy, low-risk types instead.
- Rules that react to changes. Detective rules triggered by configuration changes will only see the daily snapshot. A rule meant to flag an open security group within minutes should stay on a continuously recorded type.
- Resources that change rarely. Below four changes a day, daily recording costs more. Check the metric first.
- Managed recorders. If the recorder is set up by AWS Control Tower or an organisation-wide template, change it there; hand edits can be overwritten or flagged as drift.
- Rule evaluations, not CIs. If Cost Explorer shows rule or conformance pack evaluations as the big line, recording frequency will not help much. Remove duplicate rules first; AWS's own cost recommendations for Config cover that.
The bottom line
AWS Config charges $0.003 for every change it records, so a few high-churn resource types usually drive most of the bill. Find them with the CloudWatch metrics, switch types that change more than four times a day to daily recording, exclude types you never look at, and keep security-critical types continuous.
How LessBill can help
LessBill.org watches your AWS, Azure and GCP bills for lines like this one and flags when a governance service starts growing faster than the resources it watches. You see the cause, not just the total.
Sources
- AWS Config pricing
- AWS Config now supports periodic recording (Nov 2023)
- AWS Config simplifies usage analysis with CloudWatch metrics (Apr 2024)
- Changing the recording frequency (AWS Config docs)
- Recording resources with the AWS CLI (AWS Config docs)
- Cost optimization recommendations for AWS Config (AWS blog)
Comments
Post a Comment