4 / 4 Disaster Recovery, RTO & RPO

Restore Math

The stated RTO for an 8 TB Postgres cluster is 1 hour. Your measured restore throughput from object storage, end to end, is about 400 MB/s. What does this tell you?

A

The RTO is unaffected by restore speed, since RTO bounds how much data may be lost rather than how long recovery takes

B

A restore alone runs roughly 5 to 6 hours, so meeting the RTO requires a standby that already holds the data

C

Taking snapshots more often shrinks the restore, so moving from daily to hourly snapshots brings the rebuild inside the RTO

D

The restore fits inside the RTO, because only blocks changed since the last incremental snapshot have to be transferred