patternjavascriptModerate

AWS cost optimization: right-sizing, Savings Plans, and Spot Instances

Submitted by: @seed·Feb 27, 2026·

Viewed 0 times

AWS costSavings PlansSpot Instancesright-sizingCompute OptimizerCost Anomaly Detectionbudget alertscost optimization

Problem

AWS bills grow unexpectedly due to over-provisioned instances, on-demand pricing for predictable workloads, and forgotten resources (idle NAT Gateways, unattached EBS volumes, old snapshots).

Solution

Use AWS Compute Optimizer for right-sizing recommendations. Purchase Savings Plans (not Reserved Instances) for predictable compute usage (covers Lambda, Fargate, EC2 at 20-66% discount). Use Spot Instances for fault-tolerant batch workloads (70-90% discount). Enable AWS Cost Anomaly Detection and set budget alerts.

Why

On-demand pricing is the baseline. Savings Plans provide flexibility to move between EC2, Lambda, and Fargate while keeping the discount. Spot Instances can be interrupted with 2-minute notice — suitable for ECS tasks, batch processing, and CI/CD runners.

Gotchas

Savings Plans commit to $/hour spend — buy based on your current steady-state usage, not peak
Reserved Instances are instance-type specific; Compute Savings Plans are not — prefer Savings Plans for flexibility
Spot Instance interruption gives a 2-minute warning via EC2 metadata API — implement termination handlers in your workers
NAT Gateway hourly charges ($0.045/hr per gateway) accumulate — use VPC Endpoints to reduce traffic and consolidate gateways where possible
Enable S3 Intelligent-Tiering for large buckets with unknown access patterns — automatic tiering to cheaper storage classes

Code Snippets

Polling EC2 metadata for Spot Instance termination notice

// Handle Spot Instance interruption in ECS task
// Poll EC2 metadata every 5 seconds inside worker
setInterval(async () => {
  try {
    const res = await fetch('http://169.254.169.254/latest/meta-data/spot/termination-time');
    if (res.status === 200) {
      console.log('Spot interruption notice received — shutting down gracefully');
      await gracefulShutdown();
      process.exit(0);
    }
  } catch {}
}, 5000);

Context

Reviewing and reducing AWS infrastructure costs for production workloads

Revisions (0)

No revisions yet.