How I Reduced AWS Costs by 84%

At TechLabs, our AWS bill had crept up to $1,900/month. For a small team, that was eating significantly into our margins. Here's how I systematically reduced it to under $300.

The Problem

Like many startups, we had accumulated cloud debt: - Development environments running 24/7 - Oversized instances "just in case" - Orphaned EBS volumes and snapshots - No reserved instance strategy

Step 1: Visibility First

Before optimizing, I needed to understand where money was going. I set up:

1. . **AWS Cost Explorer** with daily granularity
2. . **Cost allocation tags** on every resource
3. . **Budgets and alerts** for each team/project

This alone revealed that 40% of our spend was on resources no one was using.

Step 2: Right-Sizing

The biggest wins came from right-sizing:

• Moved from m5.xlarge to t3.medium for most workloads (70% savings)
• Identified instances running at <10% CPU utilization
• Used AWS Compute Optimizer recommendations

Step 3: Scheduled Scaling

Development and staging environments don't need to run at night or weekends:

# Lambda function to stop dev environments at 7 PM
def stop_dev_instances(event, context):
    ec2 = boto3.client('ec2')
    instances = ec2.describe_instances(
        Filters=[{'Name': 'tag:Environment', 'Values': ['dev', 'staging']}]
    )
    # Stop instances logic here

This saved ~60% on non-production environments.

Step 4: Reserved Instances & Savings Plans

For production workloads with predictable usage, I purchased: - 1-year reserved instances for databases - Compute Savings Plans for EC2

Results

The key lesson: cloud cost optimization isn't a one-time task. It requires ongoing visibility and regular reviews.