Cloud Cost Management and Governance
What to know
Cost management begins with visibility: tagging resources by owner, environment, and cost center so spend can be attributed. Budgets and alerts catch overruns; showback/chargeback assigns costs to teams. Understand pricing models: on-demand (flexible, highest unit price), reserved/committed (discount for term commitment), and spot/preemptible (deep discount, interruptible - only for fault-tolerant workloads). Right-sizing matches instance size to actual utilization; autoscaling and scheduling (shut down non-production off-hours) reduce waste. Egress and inter-region traffic are common hidden costs.
Governance
Governance is the guardrails layer: policy as code enforcing allowed regions, instance types, and required tags; quotas and limits preventing runaway spend; approval workflows for exceptions. Governance should enable teams, not block them - prefer automated guardrails over manual review.
Traps
Questions imply spot instances suit databases or stateful workloads (they don't - interruption risk). Don't confuse cost allocation tags (attribution) with budget alerts (notification). Reserved commitments save money only if utilization is high; buying them for spiky workloads wastes money. And deleting a resource may not stop charges - orphaned storage and snapshots persist.
Worked example
A team runs batch image processing that can restart on failure and runs a few hours daily. Answer: spot/preemptible instances plus scheduling and autoscaling, with tagging for attribution. Rule of thumb: match the pricing model to the workload's interruption tolerance and utilization profile - commit only what runs steadily.