MQ

Mohd Baquir Qureshi

Technical Blog
✨
Cloud Computing

Cloud Cost Optimization Strategies

By Mohd Baquir Qureshi •
Charts showing optimization

Cloud computing was sold on the promise of "only pay for what you use." However, most companies end up paying for what they forgot to turn off. A rising discipline called FinOps has emerged simply to stop developers from accidentally bankrupting their companies on AWS.

1. Kill Zombie Infrastructure

The easiest way to save money is to turn things off.

  • Unattached EBS Volumes: When you terminate an EC2 instance, the attached block storage (EBS) does not automatically delete unless you explicitly configured it to. You could be paying thousands of dollars a month for hard drives attached to nothing.
  • Idle Load Balancers: Application Load Balancers cost money per hour, even if no traffic is hitting them. If the staging environment is down, delete the ALB.
  • Orphaned Snapshots: Automated database backups are great, but storing daily snapshots from 4 years ago is a massive waste of S3 costs. Implement aggressive lifecycle policies.

2. Right-sizing Instances

Developers love safety margins. If an app needs 2GB of RAM, they will provision an instance with 16GB "just in case."

Use AWS Compute Optimizer or GCP Recommender to find instances that have been sitting at 5% CPU utilization for the last month. Downsize them. Moving from an m5.xlarge to an m5.large immediately cuts your compute bill in half.

3. Reserved Instances & Savings Plans

If you know you are going to be running a database or a core set of web servers 24/7 for the next year, you should never pay On-Demand prices.

Committing to a 1-year or 3-year Compute Savings Plan can yield discounts of up to 72% compared to On-Demand pricing. Always buy Compute Savings Plans over specific EC2 Reserved Instances, as they offer the flexibility to change instance families later.

4. Spot Instances for Background Workers

Spot instances are spare compute capacity that AWS sells at up to a 90% discount. The catch? AWS can terminate your instance with a 2-minute warning if they need the capacity back.

You should never run your primary database on a Spot instance. However, if you have stateless Celery workers processing background queues (e.g., sending emails, resizing images), Spot instances are the perfect use case. If the instance dies, the task just goes back into the queue for another worker to pick up.

Conclusion

Cloud cost optimization is not a one-time project; it is a continuous culture. By tagging your resources properly, utilizing Spot instances, and committing to Savings Plans, you can drastically reduce your monthly burn rate.