Amazon Relational Database Service (RDS) is AWS's fully managed offering for running relational databases in the cloud. It handles the undifferentiated work -- hardware provisioning, database setup, patching, backups, scaling -- so teams can focus on their applications instead of database administration. You pick your engine, choose an instance size, and AWS takes care of the rest.
RDS supports six database engines: MySQL, PostgreSQL, MariaDB, Oracle, SQL Server, and Db2. Amazon Aurora (MySQL and PostgreSQL-compatible) is also part of the RDS family, though it runs on a different architecture under the hood.
Key Features
Automated Backups: Point-in-time recovery with configurable retention up to 35 days. RDS backs up both the database and transaction logs automatically. Need to restore to a specific second? You can, up to the last five minutes.
Multi-AZ Deployments: Two options for high availability. The classic approach maintains a synchronous standby replica in a separate Availability Zone with automatic failover. The newer Multi-AZ DB Cluster configuration uses one writer and two readable standbys across three AZs, with failover typically under 35 seconds and zero data loss.
Read Replicas: Asynchronous replicas for offloading read traffic. MySQL supports up to 15 read replicas; PostgreSQL, MariaDB, Oracle, and SQL Server also support them. Cross-region replicas are available for disaster recovery and latency reduction.
Automatic Patching: OS and engine patches applied during configurable maintenance windows. No manual intervention required.
Encryption: Data encrypted at rest with AWS KMS and in transit with SSL/TLS. Covers replicas, backups, and snapshots.
Performance Insights: A built-in monitoring tool that visualizes database load as average active sessions, sliced by wait states, SQL statements, hosts, or users. Makes it straightforward to identify bottlenecks -- whether they're CPU, lock waits, or I/O latency -- and trace them back to specific queries.
RDS Proxy: A fully managed connection pooler that sits between your application and the database. Multiplexes connections, reduces failover times by up to 66%, and handles connection surges without overwhelming the database. Particularly useful for serverless and high-concurrency workloads.
RDS Custom
Standard RDS deliberately limits OS-level access -- that's part of the managed deal. But some applications need custom configurations, specific driver installations, or native features that require root access.
RDS Custom fills this gap for Oracle and SQL Server. It gives you access to the underlying operating system and database environment while retaining managed-service benefits like automated backups, scaling, and monitoring. Oracle deployments support Bring Your Own License; SQL Server supports both License Included and Bring Your Own Media.
How RDS Differs from Self-Managed Databases on EC2
Running a database on EC2 gives you full control -- any engine version, any configuration, full OS access. The tradeoff is that you own everything: patching, backups, replication, failover, monitoring, and capacity planning.
RDS eliminates most of that operational burden. Multi-AZ failover is automatic. Backups happen on schedule with point-in-time recovery built in. Patching is handled for you. The cost per instance is higher than bare EC2, but the reduction in DBA time and operational risk often makes it cheaper overall -- especially for small-to-medium deployments.
The decision usually comes down to customization needs. If you're running a standard MySQL or PostgreSQL workload, RDS makes sense. If you need a specific engine version AWS doesn't support yet, unusual OS-level tuning, or an engine RDS doesn't offer, EC2 gives you that flexibility.
Pricing Model
RDS pricing has several components:
- Compute: Per-instance-hour billing (per-second for most engines). Choose On-Demand for flexibility or Reserved Instances for 1-3 year commitments with up to 63% savings. AWS also introduced Database Savings Plans in late 2025 -- a 1-year hourly commitment with roughly 35% savings and more flexibility than Reserved Instances.
- Storage: General Purpose (gp3) or Provisioned IOPS (io2) volumes, charged per GB-month. Storage auto-scaling is available up to a specified maximum.
- I/O: Charged per million I/O requests for Provisioned IOPS storage.
- Backup storage: Free up to the total size of your provisioned database storage. Beyond that, per GB-month.
- Data transfer: Standard AWS data transfer charges for cross-region and internet traffic.
Graviton-based instances (Arm architecture) deliver up to 40% better price-performance compared to x86 equivalents. The latest Graviton4 instances (db.m8g, db.r8g) launched in late 2025 with further improvements.
Common Use Cases
- Web and Mobile Applications: Transactional backends for websites and apps that need high throughput, automated failover, and minimal operational overhead.
- SaaS Platforms: Multi-tenant database backends with elastic storage and compute scaling. Read replicas handle reporting without impacting production.
- Enterprise Application Migration: Moving on-premises Oracle or SQL Server workloads to the cloud. RDS Custom supports the customizations legacy applications often require.
- E-Commerce: Product catalogs, orders, and inventory management with PCI-compliant security features.
- Development and Testing: Fast provisioning and teardown of database environments. SQL Server Developer Edition is available on RDS without licensing costs.
Limitations
RDS abstracts away infrastructure management, which means less control. No SSH access to the underlying OS (except through RDS Custom for Oracle and SQL Server). Engine versions are limited to what AWS supports -- and when a version reaches end of standard support, Extended Support charges kick in automatically at $0.10-0.20 per vCPU-hour. Storage limits top out at 64 TiB for most engines (256 TiB for Oracle and SQL Server with additional volumes). Connection-heavy workloads can exhaust instance memory, making RDS Proxy essential for high-concurrency scenarios.
Beyond RDS
Many organizations run RDS alongside specialized data systems. A PostgreSQL or MySQL instance on RDS handles transactional workloads, while analytical queries, full-text search, or real-time aggregations run on purpose-built engines like ClickHouse, Elasticsearch, or OpenSearch. Getting data flowing reliably between these systems -- with the right latency, consistency, and cost profile -- is where the architecture decisions get interesting.