High availability & DR

Multi-AZ, multi-Region, RTO/RPO strategies.

Intermediate30 min · lesson 10 of 15

Designing for high availability and disaster recovery is the heart of the Reliability pillar. The guiding assumption is that everything fails eventually, so you architect to absorb failures and recover within defined objectives.

Multi-AZ and multi-Region

High availability starts with multi-AZ: spread every tier across at least two Availability Zones so a data-center failure is absorbed — Auto Scaling groups across AZs behind a load balancer for compute, Multi-AZ databases for data. For protection against a whole-Region failure or for global latency, multi-Region designs replicate data and stand up capacity in a second Region, coordinated by Route 53 failover or latency routing. Multi-AZ is the default expectation for production; multi-Region is added when the business requires surviving a Region-wide event or serving users globally with local latency, and it costs and complicates more.

HA within a Region, DR across Regions
# HA (within a Region) — the production default:
# ALB → Auto Scaling group across AZ-a/b/c
# RDS Multi-AZ (synchronous standby, auto failover)
#
# DR (across Regions) — for Region-level protection:
# replicate data to a second Region (e.g. Aurora Global, S3 CRR)
# Route 53 failover routing → promote the DR Region on outage

DR strategies, RTO and RPO

Disaster-recovery strategies trade cost against recovery speed. Backup and restore is cheapest and slowest — restore from backups after an event. Pilot light keeps a minimal core (like a replicated database) always running in the DR Region, scaled up on failover. Warm standby runs a scaled-down but functional copy ready to take traffic. Multi-site active-active runs full capacity in both Regions for near-zero downtime and the highest cost. You choose based on two objectives: RTO (Recovery Time Objective — how long you can be down) and RPO (Recovery Point Objective — how much data loss is acceptable). Tighter RTO/RPO demands a more expensive strategy; match the strategy to the business requirement, not to fear.

DR strategies by cost and speed
cheaper / slower
backup & restore
restore after the event
pilot light
minimal core always on
costlier / faster
warm standby
scaled-down live copy
multi-site active-active
full capacity, near-zero RTO
Pick a DR strategy from your RTO/RPO. Tighter objectives cost more — match the strategy to the business requirement.
Untested DR is not DR
A disaster-recovery plan that has never been exercised often fails when it is finally needed — replication gaps, stale runbooks, permissions that do not work. Regularly test failover to the DR Region and measure actual RTO/RPO against your targets; the first real failover should never be during an actual disaster.