Amazon Aurora

Amazon Aurora is a fully managed relational database engine compatible with MySQL and PostgreSQL. It combines high-end commercial database performance with the simplicity and cost-effectiveness of open-source databases, delivering up to 5x MySQL and 3x PostgreSQL throughput through a cloud-native architecture that separates compute from storage.

• Aurora replicates every piece of data 6 times across 3 Availability Zones for fault tolerance and automatic self-healing.
• Supports up to 15 read replicas reading from shared storage with sub-10ms replication lag, eliminating traditional copy overhead.
• Offers provisioned and serverless deployment modes for predictable and variable workloads.
• Includes advanced features like Global Database for multi-region replication, Backtracking for instant point-in-time recovery, and Fast Cloning.
• Failover occurs in under 30 seconds with automatic promotion of a read replica to primary.

Architecture

Unlike traditional databases that write to local EBS volumes, Aurora uses a log-structured distributed storage system independent of compute instances.

• 6 Copies Across 3 AZs: Every write is automatically replicated 6 times across 3 Availability Zones.
• Fault Tolerance: Tolerates loss of up to 2 copies without affecting write availability and up to 3 copies without affecting read availability.
• Self-Healing: Continuously scans data blocks, detects errors, and repairs them automatically from other copies.
• Shared Storage Volume: Up to 15 read replicas read from the same underlying storage, eliminating the data copy overhead that causes replication lag in traditional systems.

Cluster Architecture

An Aurora cluster consists of a writer instance, multiple reader instances, and a shared cluster volume spanning multiple Availability Zones.

• Primary Instance (Writer): Handles all read and write operations; exactly one per cluster.
• Reader Instances (Read Replicas): Handle read-only operations; up to 15 per cluster.
• Automatic Failover: If the primary fails, Aurora automatically promotes a reader to primary in under 30 seconds.
• Multi-Master Cluster: All instances have read and write capabilities for distributed write operations across zones.

Deployment Modes

• Aurora Provisioned: Specify a fixed instance class (e.g., db.r6g.large); best for predictable steady-state workloads. Scale manually by changing instance type or use Auto-Scaling for read replicas.
• Aurora Serverless v2: Define a capacity range (min/max Aurora Capacity Units). Automatically scales compute and memory in fractions of a second based on actual traffic. Best for variable workloads, dev/test, and SaaS applications with unpredictable traffic spikes. Pay per ACU-second consumed.

Advanced Features

• Aurora Global Database: Spans multiple AWS regions with typical replication latency under 1 second. Enables local reads for geographically distributed users and disaster recovery with secondary region promotion in under 1 minute.
• Backtracking: Instantly rewinds the database to any prior point in time without restoring a backup snapshot. Invaluable for fixing bad deployments or accidental deletions.
• Fast Cloning: Creates a new cluster from an existing one in minutes regardless of data size using copy-on-write protocol. The clone shares storage blocks with the original until changes are made, avoiding double storage costs. Ideal for spinning up full-size production clones for testing or analytics.

Aurora vs RDS (MySQL/PostgreSQL)

Aurora Pricing

• Compute: Provisioned: hourly rate per instance size. Serverless v2: per ACU-hour billed per second.
• Storage: Pay per GB-month for data stored; auto-expanding with no pre-allocation required.
• I/O Rate — Aurora Standard: Pay per million I/O requests; best for low-I/O applications.
• I/O Rate — Aurora I/O-Optimized: Higher storage and instance cost with zero I/O charges; best for I/O-intensive workloads such as payment processors.