-
Book Overview & Buying
-
Table Of Contents
The Platform Engineer's Handbook
By :
Objective: Understand how to combine Committed Use Discounts (CUDs), on-demand, and spot instances to optimize cloud cost without materially impacting performance. At NewTech, Maria's team committed 20 vCPU and 40GB RAM of baseline capacity to a 3-year CUD. AWS gave them 45% off the list price.
For workloads exceeding the baseline, they used a mix of On-demand instances (for stability) and Spot instances (for cost savings).
The outcome was 60% total savings compared to running everything on on-demand, with only a 1% increase in latency due to spot interruptions.
Question: If NewTech had committed more than their true baseline (for example, 35 vCPU instead of 20 vCPU), what financial or operational risks might they have introduced?
Cost optimization is not a once-and-done activity. It's a continuous practice with appropriate guardrails to prevent teams from accidentally or intentionally consuming unlimited resources...