RPO · Recovery Point Objective
The most data a service accepts losing after a failure, measured in time. For training it is the age of the newest checkpoint you can recover inside the failure domain you design for.
Current numbers
about 220 Gbit/s payloadIllustrative 100 TB decimal bulk copy in one hour; about 220 Gbit/s payload, independent of the RPO test
15 min RTO / 5 min RPO (AWS example tier)AWS Platinum resilience-tier example for mission-critical workloads; you choose your own RTO and RPO