Skip to main content

Downsizing an Underutilized EC2 Instance

Overview​

This tutorial demonstrates how to reduce costs by downsizing an underutilized EC2 instance. We'll identify an instance with consistently low CPU and network usage, then resize it to a smaller instance type to achieve approximately 50% cost savings.

Target Instance: i-0cd15bcce35c73ac9 Original Type: t3.micro ($0.0104/hour) New Type: t3.nano ($0.0052/hour) Monthly Savings: ~$37

Prerequisites​

  • AWS Console access with EC2 permissions
  • An underutilized EC2 instance to remediate

Steps​

1. Navigate to EC2 Instances​

Open the AWS Console and navigate to the EC2 service. Select "Instances" from the left sidebar to view all running instances.

EC2 Instances Console

2. Locate the Underutilized Instance​

Use the search filter to find your target instance by Instance ID. Select the instance to view its details.

3. Review Monitoring Metrics​

Click on the "Monitoring" tab to view CloudWatch metrics for the instance. Examine CPU utilization and network traffic over the past few hours.

Monitoring Tab

4. Analyze CPU Utilization​

Look at the CPU utilization graph. If the instance shows consistently low CPU usage (e.g., below 5%), it's a good candidate for downsizing.

CPU Metrics

In this example, the maximum CPU utilization is only 0.53%, indicating significant underutilization.

5. Stop the Instance​

Before changing the instance type, you must stop the instance. Select the instance, click "Instance state" → "Stop instance".

Stop Confirmation

Click "Stop" to confirm. Wait for the instance state to change to "Stopped".

6. Change Instance Type​

Once stopped, select the instance and click "Actions" → "Instance settings" → "Change instance type".

Select Instance Type

Select a smaller instance type (e.g., t3.nano) from the dropdown. The console will show the cost comparison:

  • t3.nano: $0.0052/hour, 512 MiB memory
  • t3.micro: $0.0104/hour, 1024 MiB memory

This represents a 50% cost reduction.

7. Apply the Change​

Click "Apply" to change the instance type.

Apply Change

You should see a success notification confirming the instance type has been changed.

8. Restart the Instance​

Select the instance and click "Instance state" → "Start instance" to restart it with the new instance type.

Instance Running

Wait for the instance to reach the "Running" state. Verify that the instance type column shows the new type (t3.nano).

9. Verify the Change​

After the instance is running, check:

  • Instance type is correctly set to t3.nano
  • Instance has a new public IP (if using dynamic IPs)
  • Applications running on the instance are functioning normally

Cost Impact​

Before: t3.micro at $0.0104/hour = $7.59/month (730 hours) After: t3.nano at $0.0052/hour = $3.80/month (730 hours) Monthly Savings: $3.79 (50% reduction) Annual Savings: $45.48

For multiple underutilized instances, these savings can add up significantly.

Alternative Approaches​

Option 1: Schedule Instance Downtime​

If the instance is only needed during business hours, consider using AWS Instance Scheduler to automatically stop instances during off-hours. This can save up to 70% on compute costs.

Option 2: Switch to ARM-based Instances​

Consider migrating to ARM-based Graviton instances (t4g family) for up to 20% additional cost savings compared to x86 instances. Note: This requires application compatibility with ARM architecture.

Option 3: Use Auto Scaling with Scheduled Scaling​

For predictable workload patterns, configure Auto Scaling with scheduled scaling policies to automatically adjust capacity based on time of day or day of week.

Option 4: Move to Serverless​

For very low utilization workloads, consider migrating to AWS Lambda or Fargate to pay only for actual compute time used rather than idle instance hours.

Best Practices​

  • Monitor After Resizing: Watch CloudWatch metrics for 24-48 hours after downsizing to ensure the smaller instance type handles the workload adequately.
  • Set CloudWatch Alarms: Create alarms for high CPU utilization (>80%) to be notified if the instance becomes undersized.
  • Review Regularly: Perform monthly reviews of instance utilization to identify opportunities for further optimization.
  • Test Before Production: Always test instance type changes in non-production environments first to validate performance.

Rollback Procedure​

If you need to restore the original instance type:

  1. Stop the instance
  2. Change instance type back to t3.micro (or original type)
  3. Start the instance
  4. Verify application performance

Additional Resources​