AZ-305 ยท Backup and disaster recovery
RTO, RPO and choosing a recovery strategy
Defining RTO and RPO per workload, mapping them to backup vs replication vs active-active, cost trade-offs, the Well-Architected reliability view of DR.
Key points
- RPO: acceptable data loss, in time.
- RTO: acceptable time until the workload runs again.
- Patterns: backup and restore (cheapest, longest RTO), cold standby (deployed at failover), warm standby (reduced copy), active-active (fastest, costliest).
- Synchronous replication: zero loss, more latency; asynchronous: some loss.
- Well-Architected: targets from criticality tiers, tested DR plan with drills, separate failback.
Source: https://learn.microsoft.com/en-us/azure/well-architected/reliability/disaster-recovery
Practice questions
What does a recovery point objective (RPO) of 1 hour tell you about a workload?
- The service must be running again within 1 hour of an outage
- At most 1 hour of data changes may be lost after a disaster
- Backups must be retained for at least 1 hour
- A DR drill must finish within 1 hour
Show answer
Answer: B. At most 1 hour of data changes may be lost after a disaster
RPO is the maximum acceptable data loss expressed as time, so at most the last hour of changes may be lost. The time allowed to bring the service back is the RTO, not the RPO.
A workload must recover from a regional outage faster than today. Which changes shorten the RTO?
- Run a reduced-capacity copy of the compute tier in the secondary region
- Automate failover with declarative, idempotent scripts and pipelines
- Extend backup retention from 30 days to 90 days
- Pre-create the VNets, subnets and firewall rules in the secondary region
- Change the backup vault from geo-redundant to locally redundant storage
Show answer
Answer: A. Run a reduced-capacity copy of the compute tier in the secondary region, B. Automate failover with declarative, idempotent scripts and pipelines, D. Pre-create the VNets, subnets and firewall rules in the secondary region
Pre-deploying compute (warm standby), replicating the network topology ahead of time and automating the failover all cut the time needed to get running in the secondary region. Longer backup retention changes how far back you can restore, not how fast; switching the vault to LRS removes cross-region copies and makes regional recovery harder.
Lowering a workload's RPO automatically lowers its RTO as well.
Show answer
Answer: False
False. RPO and RTO are independent targets. Replicating every few minutes gives a small RPO, but if the secondary environment still has to be deployed and configured by hand, the RTO stays long.
Practice 8 questions on this topic in Bitzu
Short lessons, instant explanations and spaced review of what you get wrong. Free on iOS and Android.
