Skip to main content

Description

Identify Amazon EC2 instances with minimal activity over the past 14 days and address them to reduce unnecessary costs. When a linked New Relic host is available, this policy combines CloudWatch CloudWatch: CPU, 14-Day and CloudWatch: CPU Max, 14-Day with New Relic memory, disk read utilization, and disk write utilization plus 14-day network metrics. Otherwise, it uses the existing CloudWatch CPU and network metrics. Instances are classified as idle if they meet the applicable criteria:

  • CPU utilization below 5%
  • Maximum CPU utilization below 15%
  • Memory utilization below 20%
  • Disk read utilization below 5%
  • Disk write utilization below 5%
  • 14-day network I/O below 100 MB

Rationale​

Idle EC2 instances consume resources and incur costs without contributing meaningfully to operations. Adding maximum CPU utilization as a criterion ensures instances experiencing brief spikes in usage are not prematurely identified as idle. This approach improves accuracy in cost-saving measures while maintaining operational efficiency. Addressing idle instances reduces waste and optimizes the AWS environment.

Audit​

This policy evaluates an AWS EC2 instance over the last 14 days using CPU, network, and memory metrics.

Memory is evaluated in this order:

  • If New Relic Host is present, use New Relic memory and disk utilization metrics.
  • Otherwise, use CloudWatch (Agent): Memory Used, 14-Day.
  • If that metric is empty, use Nagios: Memory Utilization.
  • If all memory metrics are empty, fall back to CPU and network only.

The instance is marked as INCOMPLIANT when all of these baseline conditions are true:

  • CloudWatch: CPU, 14-Day is less than 5%.
  • CloudWatch: CPU Max, 14-Day is less than 15%.
  • CloudWatch: Network In, MB, 14-Day is less than 100 MB.
  • CloudWatch: Network Out, MB, 14-Day is less than 100 MB.

And one of these metric paths applies if present:

  • New Relic Host is present, New Relic Host: Memory Used, 14-Day is less than 20%, New Relic Host: Disk Read Utilization is less than 5%, and New Relic Host: Disk Write Utilization is less than 5%.
  • CloudWatch (Agent): Memory Used, 14-Day is less than 20%.
  • Nagios: Memory Utilization is less than 20%.

The instance is marked as INAPPLICABLE if it is not currently running or has been running for less than 14 days.

The instance is marked as UNDETERMINED if any required CPU or network metric is empty, or if New Relic Host is present but a required New Relic memory or disk metric is empty.