Learn Cloud Computing - FinOps, Cost Management & Cloud Optimization

#Introduction

After episode 16, where you created infrastructure as fast as writing code, a question often realized too late arises: who pays for all of it? The ease of provisioning in the cloud is a double-edged sword — the very thing that lets a team build environments in minutes can also balloon a bill in days.

This episode covers FinOps: a discipline that unites engineering, finance, and business in managing cloud costs. You'll learn the most impactful optimization strategies — right-sizing, auto-shutdown, spot instances, savings plans, and storage lifecycles — then close with budgets and alerts that keep costs under control.

#Main Discussion

#What Is FinOps

FinOps is a cloud cost management discipline that unites three parties: the engineering team that uses resources, the finance team that pays the bills, and the business team that defines service value. Its core philosophy: cost isn't finance's business alone — engineers carry direct responsibility for the resources they create.

The most fitting analogy is a restaurant kitchen. Chefs choose and use ingredients (engineering), the budget cook watches spending (finance), and the owner decides which dishes sell (business). A healthy restaurant doesn't hand all kitchen control to the accounting department — the chef knows the price of every ingredient they use, because only they know which ingredients are essential and which are just habit.

The FinOps journey rotates through three phases:

1. Inform — visibility: knowing how much is being spent and by whom, especially through tagging and cost allocation.
2. Optimize — lowering costs: right-sizing, removing unused resources, and using available discounts.
3. Operate — keeping costs under control: budgets, alerts, and periodic reviews as routine.

#Cost Optimization Strategies

The four following strategies account for most savings in the field. The order is deliberate: from the most often forgotten to the one requiring the most planning.

1. Right-sizing. Over-provisioned instances are the most common waste — people tend to pick big sizes "just to be safe". Use the metric data from episode 15 (CPU, memory, network) to choose the right size, and measure not from momentary peak load but from weekly trends. An instance averaging only 15 percent CPU deserves to be lowered one or two levels.

2. Auto-shutdown staging and dev environments. Non-production environments don't need to run 24/7. Schedule them off at night and on weekends — AWS Instance Scheduler, GCP via Cloud Scheduler, Azure with shutdown schedules via an Automation account. This single step can save up to 60 percent of non-production compute costs without changing application behavior at all.

3. Spot and Savings Plans. For workloads tolerant of interruption — batch jobs, rendering, CI — spot instances can save up to 90 percent because the cloud may reclaim their capacity. For stable production loads, Savings Plans or reserved instances give 30 to 60 percent discounts in exchange for a 1- to 3-year commitment. Use spot for the flexible, savings plans for the stable.

4. Storage lifecycle to archive. Rarely accessed data doesn't need to live in the expensive standard tier. Create lifecycle rules so objects automatically move from the standard tier to infrequent access, then to archive after a certain period:

#Seeing and Holding Costs: Budgets and Alerts

Optimization without visibility is like driving without a speedometer. The cloud provides tools to see and hold back costs:

• AWS — Cost Explorer for trend analysis, and Budgets to set limits with alerts and automation.
• GCP — Cloud Billing reports for analysis, and budgets & alerts for thresholds.
• Azure — Cost Management + Billing for analysis, and Azure Advisor for optimization recommendations.

Here's an example of viewing last month's costs per service via the Cost Explorer API:

aws ce get-cost-and-usage \
  --time-period Start=2026-07-01,End=2026-08-01 \
  --granularity MONTHLY \
  --metrics "UnblendedCost" \
  --group-by Type=DIMENSION,Key=SERVICE

The aws ce get-cost-and-usage command above returns a cost grouping per service:

{
  "ResultsByTime": [
    {
      "TimePeriod": {
        "Start": "2026-07-01",
        "End": "2026-08-01"
      },
      "Groups": [
        {
          "Keys": ["Amazon Elastic Compute Cloud"],
          "Metrics": {
            "UnblendedCost": { "Amount": "412.50", "Unit": "USD" }
          }
        }
      ]
    }
  ]
}

These numbers are the basis for decisions: which service dominates the bill, whether the pattern is reasonable, and which optimization strategy makes the most sense to apply.

A good budget doesn't just record — it sends alerts early. Here's an example monthly budget with an alarm at the 80 percent threshold:

{
  "BudgetName": "monthly-production",
  "BudgetType": "COST",
  "BudgetLimit": { "Amount": "1000", "Unit": "USD" },
  "TimeUnit": "MONTHLY",
  "CostFilters": {},
  "CostTypes": {
    "IncludeCredit": false,
    "IncludeRefund": false
  }
}

{
  "Notifications": [
    {
      "Notification": {
        "NotificationType": "ACTUAL",
        "ComparisonOperator": "GREATER_THAN",
        "Threshold": 80,
        "ThresholdType": "PERCENTAGE"
      },
      "Subscribers": [
        { "SubscriptionType": "EMAIL", "Address": "devops@company.id" }
      ]
    }
  ]
}

When spending passes 80 percent of the budget, the email arrives early — well before the bill becomes a surprise at the end of the month. With both files, the budget is created via aws budgets create-budget in one step.

#Comparing the Big 3 Cost Services

As in previous episodes, the concepts are identical — visibility, optimization, and operationalization — only the service names differ. What distinguishes a team's maturity level is its discipline: tags managed properly, scheduled cost reviews, and budgets enforced through automation.

#Conclusion

In this episode 17 you learned the economics side of the cloud: FinOps — a collaboration of engineering, finance, and business with three phases: inform, optimize, operate; four optimization strategies — right-sizing with metric data, auto-shutdown for staging and dev, spot and savings plans for discounts, and storage lifecycles to archive for cold data; budgets and alerts — setting limits and automating follow-up; and a tool comparison across AWS, GCP, and Azure.

At this point you can build, secure, monitor, automate, and control costs within one cloud. The next question arises when an organization isn't satisfied with a single provider: how do you run workloads across two clouds at once, or connect a cloud with your own on-premise data center? The next episode, Hybrid Cloud & Multi-Cloud Connectivity, opens the world of networking beyond the boundaries of a single provider.