Optimization starts with evidence about demand, utilization, reliability, and operating work. Removing idle resources can help, but deleting a recovery dependency without checking its purpose can violate the design. Right-sizing uses measurements and tests rather than assuming the smallest instance is adequate.
Purchasing models address different uncertainty. A steady eligible baseline may suit an appropriate commitment after current terms are reviewed. Interruptible capacity can fit restartable work with checkpointing, but it doesn't suit every database or latency-sensitive request. Storage transitions also need retrieval and retention analysis. Keep availability, security, and recovery constraints in the comparison. In the edge task, an apparently idle standby is still required for the accepted failure scenario, so its low normal utilization is expected.