RDS Multi-AZ and Read Replicas

Production databases need to survive hardware failures and handle growing read traffic. Amazon RDS provides two complementary features to address these requirements: Multi-AZ deployments for high availability and read replicas for horizontal read scalability.

---

Multi-AZ Deployments

How It Works

When you enable Multi-AZ, RDS automatically provisions and maintains a synchronous standby replica in a different Availability Zone within the same Region.

1. Every write to the primary is synchronously replicated to the standby before RDS acknowledges the commit to your application
2. The standby is not accessible for read traffic — it exists solely for failover
3. If the primary fails, RDS performs an automatic failover to the standby, typically completing within 60–120 seconds

What Triggers a Failover?

Event	Automatic Failover?
Primary AZ outage	Yes
Primary host failure	Yes
Primary instance unresponsive	Yes
Storage failure on primary	Yes
Manual reboot with failover	Yes (you initiate)
DB engine patching	Yes (if maintenance window triggers reboot)
Scaling instance class	Brief downtime, not a failover

DNS and the Endpoint

RDS provides a single DNS endpoint for your database. During failover, RDS updates the DNS record to point to the new primary. Your application does not need to change its connection string — it just needs to handle a brief disconnection and reconnect.

mydb.cxyz123abc.us-east-1.rds.amazonaws.com → Primary (AZ-a)
                                     failover → Standby promoted (AZ-b)

Multi-AZ Pricing Impact

• You pay for two instances (primary + standby), so the cost is roughly double a single-AZ deployment
• There is no additional charge for the data replication itself
• Automated backups are taken from the standby to reduce I/O impact on the primary

---

Multi-AZ DB Cluster (Aurora-Style for RDS)

In 2022, AWS introduced the Multi-AZ DB Cluster deployment for MySQL and PostgreSQL. This creates a cluster with:

• One writer instance
• Two reader instances in separate AZs
• Each reader can serve read traffic (unlike the classic standby)
• Uses a transaction log-based replication with typically < 100 ms lag

Feature	Classic Multi-AZ	Multi-AZ DB Cluster
Standby instances	1 (not readable)	2 (readable)
Failover time	60–120 s	~35 s
Read scaling	No	Yes
Engines	All six	MySQL 8.0, PostgreSQL 13+

---

Read Replicas

How They Work

Read replicas use the engine's native asynchronous replication to stream changes from the primary to one or more replicas. Replicas are read-only and can serve SELECT queries, reducing the load on the primary.

Key Characteristics