DEV Community

Cover image for Day19: You’re Probably Overpaying for Cloud — Here’s How to Fix It 👇
Ibrahim S
Ibrahim S

Posted on

Day19: You’re Probably Overpaying for Cloud — Here’s How to Fix It 👇

#ibbus #100daysoffinops #cloudoptimization #finopssi

Cloud costs can grow fast if you’re not paying attention.Most waste comes from common, fixable issues.

  1. Right-Sizing Basics
  2. CPU & Memory Utilization Analysis
  3. Auto-Scaling Fundamentals
  4. Spot Instances / Preemptibles
  5. Container Cost Considerations
  6. Kubernetes Cost Visibility
  7. Serverless Cost Patterns
  8. Idle Resources Detection
  9. Environment-Based Optimization (Dev / Test / Prod)

Right-Sizing Basics

Right-sizing means using the correct size of resources—not too big, not too small.

Many virtual machines are overpowered “just in case.” If your app only uses 20% of the CPU, you’re paying for unused power.

  • Check actual CPU and memory usage
  • Downgrade oversized VMs

Scale up only when needed

  • Lower cost
  • Same performance

CPU & Memory Utilization Analysis

This is about measuring before changing.

  • Average CPU usage
  • Peak usage times
  • Memory consumption

If CPU stays under 30% most of the time, your instance is probably too large.

Analyze at least 7–14 days of data to avoid bad decisions.

Auto-Scaling Fundamentals

Auto-scaling automatically adds or removes resources based on demand.

Instead of running 10 servers all day:

  • Run 3 during low traffic
  • Scale to 10 during peak hours
  1. Pay only when traffic increases
  2. Better reliability
  3. No manual intervention

Spot Instances / Preemptibles

Spot (or preemptible) instances are unused cloud capacity sold at a discount—often 70–90% cheaper.

  • Batch jobs
  • CI/CD runners
  • Data processing
  • Non-critical workloads
  • These instances can be stopped at any time, so don’t use them for critical services.

Container Cost Considerations

Containers make deployment easy—but they can hide waste.

  • Requesting more CPU/memory than needed
  • Too many idle containers
  • No limits set
  1. Set resource requests and limits
  2. Monitor container-level usage
  3. Avoid overpacking nodes

Kubernetes Cost Visibility

  • Kubernetes doesn’t show costs by default.
  • Without visibility, you don’t know:
  • Which namespace costs the most
  • Which team is overspending
  • Which pods are idle
  1. Cost monitoring tools
  2. Labels for teams/projects
  3. Regular usage reviews
  4. Visibility = control

Serverless Cost Patterns

Serverless sounds cheap—but bad design can get expensive.

  • Number of executions
  • Execution time
  • Memory allocated
  1. Functions running too long
  2. Over-allocated memory
  3. High request frequency

Optimize by:

  1. Reducing execution time
  2. Right-sizing memory
  3. Caching results

Idle Resources Detection

Idle resources are the biggest source of cloud waste.

  • VMs running with no traffic
  • Unused load balancers
  • Old snapshots and disks
  • Forgotten test environments

If it hasn’t been used in 30 days, question why it exists.

Environment-Based Optimization (Dev / Test / Prod)

Not all environments need the same power.

Dev & Test:

  • Smaller instances
  • Auto shutdown at night
  • Spot instances

Production:

  • High availability
  • Auto-scaling
  • Performance-focused

Optimization Review

Optimization is not a one-time task.

Create a habit:

  • Monthly cost reviews
  • Usage dashboards
  • Alerts for spikes
  • Continuous right-sizing

Cloud cost optimization works best when it’s ongoing, not reactive.

Top comments (0)