High Availability and Read Scaling
Explore key Amazon RDS deployment patterns to ensure database high availability and efficiently scale read traffic. Understand the differences between single-AZ, multi-AZ deployments, multi-AZ clusters, and read replicas. Learn how to manage failover, read routing, and replicate data to optimize performance and minimize downtime.
Once the right instance class and storage configuration are in place, the next architectural decision determines how the database survives failures and absorbs growing read demand. A single database instance running in one Availability Zone is both a single point of failure and a throughput ceiling for read-heavy applications. This lesson separates two concerns that are often conflated but operate independently. High availability keeps the database running when the infrastructure fails. Read scaling distributes SELECT traffic across multiple instances so the primary is not overwhelmed. Amazon RDS addresses these concerns through four deployment patterns that learners must compare precisely: single-AZ, multi-AZ DB instance deployment, multi-AZ DB cluster, and read replicas. Each pattern makes different trade-offs around cost, failover speed, data durability, and whether standby or replica instances can serve read traffic. Tying all four together is the endpoint and connection routing strategy, because the application must know exactly where to send writes and reads during both normal operation and failover events. Misconfiguring the application connection after restore can extend downtime even after the restored database is available.
Single-AZ vs. multi-AZ DB instance
A single-AZ deployment runs one DB instance in one Availability Zone with no automatic failover. If the instance or the AZ fails, the database is unavailable until manual intervention restores it. This pattern suits development, testing, or noncritical workloads where cost matters more than uptime.
A multi-AZ DB instance deployment changes the resilience picture entirely. RDS provisions a
Attention: The standby in a multi-AZ DB instance deployment does not serve any read traffic. It exists solely for failover and durability. This is the most common misconception on AWS exams and in architecture reviews.