AWS RDS: AI-Optimized Technical Reference

Configuration That Actually Works in Production

Database Engine Selection Criteria

• Aurora: 2x cost of regular RDS, fast failover (<35s), auto-scaling storage
  • Avoid: Aurora Serverless v1 (broken), use v2 if needed
  • Use case: Critical workloads where cost < reliability
• PostgreSQL: Recommended choice, handles JSON well, stable performance
  • Avoid: Versions 12-13 (memory leak issues under heavy load)
  • Recommended: Version 16+ (fixed parallel query issues)
• MySQL: Adequate for most applications
  • Avoid: Version 8.4 (breaks ORMs, replication issues)
  • Stick with: Version 8.0 for stability
• Oracle: Bulletproof but expensive (licensing costs exceed infrastructure)
• SQL Server Express: 10GB limit reached faster than expected

Instance Class Configuration

• Production: Use r6i instances minimum (r7i costs 20% more for marginal gains)
• Never use in production: t3/t4g instances (burstable = unpredictable performance)
• Connection limits: PostgreSQL defaults to 100 connections on small instances

Storage Configuration

• Default choice: gp3 (20% cheaper than gp2, adequate performance)
• Avoid: io2 unless actually hitting IOPS limits (expensive, most apps don't need 256k IOPS)
• Never use: Magnetic storage (deprecated, extremely slow)

High Availability Setup

• Single-AZ: Development only (10-15 minute outages during failures)
• Multi-AZ: Production standard (60-120 second failover, doubles cost)
• Multi-AZ DB Clusters: Premium option (<35 second failover, readable standbys, MySQL/PostgreSQL only)

Real Cost Structure

Actual Production Costs vs Marketing

• Marketed: db.t4g.micro at $13/month
• Reality: Medium production setup = $440/month
  • db.r6i.large: $165/month base
  • Multi-AZ: Double to $330/month
  • 500GB gp3 storage: $60/month
  • Backup storage: $50/month
  • Data transfer charges: $200+ unexpected

Cost Reduction Strategies

• Reserved Instances: 40% savings with 1-year commitment
• gp3 migration: 20% immediate storage cost reduction
• Backup retention optimization: 7-14 days typical vs 35-day maximum
• Same-AZ deployment: Reduces cross-AZ data transfer costs

Critical Failure Modes and Solutions

Performance Bottlenecks

• Connection exhaustion: Mandatory connection pooling (pgbouncer or RDS Proxy)
• gp3 IOPS burst limits: Performance degrades during high-load periods
• Cross-AZ latency: 1-2ms added to every query with Multi-AZ
• Backup windows: I/O performance drops during automated backups

Migration Failure Scenarios

• Same engine migrations: 2-week plan becomes 6-week reality
  • Character encoding differences break applications
  • Default parameter variations cause failures
  • Connection string changes require coordinated deployments
• Cross-engine migrations: 6+ month timeline minimum
  • Schema Conversion Tool: 80% automatic, 20% manual (consumes 80% of time)
  • Stored procedures require complete rewriting
  • ORM compatibility testing essential

Security Configuration Errors

• Network misconfiguration: Never allow internet access to RDS instances
• Encryption timing: Cannot encrypt existing databases without migration
• Access control: IAM database authentication complex to implement, most use Secrets Manager

Resource Requirements and Constraints

Time Investment Reality

• Lift-and-shift: Plan 2 weeks, budget 6 weeks for debugging
• Engine migration: Minimum 6 months for complex schemas
• Production setup: Allow 1-2 weeks for proper configuration and testing

Expertise Requirements

• Database administration: Still needed for parameter tuning, monitoring
• AWS networking: VPC, security groups, subnet configuration
• Migration planning: Schema analysis, compatibility testing, rollback procedures

Breaking Points and Limits

• Storage: 128TB maximum (vs 4TB Azure limit, 64TB Google limit)
• Connections: Instance-dependent, requires pooling for scale
• Backup retention: 35-day maximum, costs accumulate rapidly
• Cross-region replication: 500ms+ lag, unsuitable for real-time requirements

Decision Criteria for Alternatives

RDS vs EC2 Database

Choose RDS when:

• Team lacks dedicated database administration expertise
• Compliance requirements favor managed services
• Business can absorb 40% cost premium for operational simplicity

Choose EC2 when:

• Custom extensions or kernel tuning required
• Cost constraints override operational complexity
• Full control over maintenance windows necessary

Aurora vs Standard RDS

Choose Aurora if:

• Budget allows 2x cost increase
• Storage auto-scaling eliminates capacity planning
• Sub-60 second failover times critical

Choose Standard RDS if:

• Cost optimization prioritized
• Predictable storage growth patterns
• Vendor lock-in concerns exist

Critical Warnings and 2025 Updates

Service Deprecations

• Performance Insights: Discontinued November 30, 2025
• Replacement: CloudWatch Database Insights (higher cost, reduced functionality)
• Instance types: t2/m4/r4 being phased out
• Database versions: MySQL 5.7 support ending, PostgreSQL 12/13 memory issues

Hidden Operational Costs

• Data transfer: Cross-AZ charges accumulate unexpectedly
• Backup storage: Scales with database size and retention period
• Maintenance downtime: Blue/green deployments double costs during migration
• Parameter changes: Most require instance restarts

Production Monitoring Essentials

• CPU utilization (standard)
• Database connections (prevents exhaustion)
• Read/write IOPS (identifies storage bottlenecks)
• Replication lag on read replicas (data consistency)
• Backup duration (indicates storage health)

Compliance and Audit Considerations

Certification Coverage

• AWS provides: SOC, PCI, HIPAA compliance frameworks
• Customer responsibility: Proper configuration implementation
• Audit requirements: Encryption key rotation logs, CloudTrail for KMS events

Security Best Practices

• Private subnet deployment mandatory
• Security groups: Default deny, specific port allowances
• SSL/TLS: Enabled by default but application must utilize
• Access control: Database passwords in Secrets Manager preferred over IAM authentication

This reference provides decision-support data for RDS implementation while preserving all operational intelligence from real-world production experience.