Evaluate architecture changes by unit cost, demand shape, transfer path, interruption tolerance, and recovery requirements.
1.Unused unattached EBS volumes exist
2.Stable 24/7 compute baseline measured for a year
3.Fault-tolerant batch jobs can checkpoint
4.Private workloads move large S3 data through NAT
5.Objects are rarely read after 90 days
6.A workload needs two AZs to meet availability
7.A cheaper node makes the HPC job run three times longer
8.Nonproduction instances run all weekend unused
9.Database is sized for a rare read burst
10.A commitment has low utilization after workload retirement