Understanding rightsizing recommendations calculations

This section provides an overview of the savings calculations that are used in your rightsizing recommendations algorithms.

Consolidated billing family

To identify all instances for all accounts in the consolidated billing family, rightsizing recommendations look at the usage for the last 14 days for each account. If the instance was stopped or terminated, we remove it from consideration. For all remaining instances, we call CloudWatch to get maximum CPU utilization data, memory utilization (if enabled), network in/out, local disk input/ output (I/O), and performance of attached EBS volumes for the last 14 days. This is to produce conservative recommendations, not to recommend instance modifications that could be detrimental to application performance or that could unexpectedly impact your performance.

Determining if an instance is idle, underutilized, or neither

We look at the maximum CPU utilization of the instance for the last 14 days to make one of the following assessments:

If the instance isn't idle or underutilized, we don't generate any recommendations.

Generating modification recommendations

Recommendations use a machine learning engine to identify the optimal Amazon EC2 instance types for a particular workload. Instance types include those that are a part of AWS Auto Scaling groups.

The recommendations engine analyzes the configuration and resource usage of a workload to identify dozens of defining characteristics. For example, it can determine whether a workload is CPU-intensive or whether it exhibits a daily pattern. The recommendations engine analyzes these characteristics and identifies the hardware resources that the workload requires.

Finally, it concludes how the workload would perform on various Amazon EC2 instances to make recommendations for the optimal AWS compute resources that the specific workload.

Savings calculation

We first examine the instance running in the last 14 days to identify whether it was partially or fully covered by an RI or Savings Plans, or running On-Demand. Another factor is whether the RI is size-flexible. The cost to run the instance is calculated based on the On-Demand hours and the rate of the instance type.

For each recommendation, we calculate the cost to operate a new instance. We assume that a size-flexible RI covers the new instance in the same way as the previous instance if the new instance is within the same instance family. Estimated savings are calculated based on the number of On-Demand running hours and the difference in On-Demand rates. If the RI isn't size-flexible, or if the new instance is in a different instance family, the estimated savings calculation is based on whether the new instance had been running during the last 14 days as On-Demand.

Cost Explorer only provides recommendations with an estimated savings greater than or equal to $0. These recommendations are a subset of Compute Optimizer results. For more performance-based recommendations that might result in a cost increase, see Compute Optimizer.

You can choose to view saving with or without consideration for RI or Savings Plans discounts. Recommendations consider both discounts by default. Considering RI or Savings Plans discounts might result in some recommendations showing a savings value of $0. To change this option, see Using your rightsizing recommendations.