Amazon RDS - Comprehensive Guide

🗄️ Tổng quan về RDS

RDS là gì?

Relational Database Service: Managed database service
Multiple engines: MySQL, PostgreSQL, MariaDB, Oracle, SQL Server, Aurora
Automated operations: Backup, patching, monitoring, scaling
High availability: Multi-AZ deployments

RDS vs Self-managed Database

Feature	RDS	Self-managed EC2
Setup	Minutes	Hours/Days
Patching	Automated	Manual
Backup	Automated	Manual setup
Monitoring	Built-in	Setup required
Scaling	Push-button	Manual
Cost	Higher per hour	Lower per hour, higher ops cost

🏗️ Database Engines

1. Amazon Aurora

MySQL/PostgreSQL compatible
Performance: 5x faster than MySQL, 3x faster than PostgreSQL
Storage: Auto-scaling from 10GB to 128TB
Replicas: Up to 15 read replicas
Multi-master: Write scaling

2. MySQL

Versions: 5.6, 5.7, 8.0
Storage: gp2, gp3, io1, io2
Max storage: 64TB
Read replicas: Up to 5

3. PostgreSQL

Versions: 11, 12, 13, 14, 15
Extensions: PostGIS, pg_stat_statements
Logical replication: Cross-region
Performance Insights: Query analysis

4. MariaDB

MySQL compatible: Drop-in replacement
Storage engines: InnoDB, MyRocks
Thread pooling: Better concurrency

5. Oracle

Editions: Standard, Enterprise
Licensing: BYOL or License Included
Features: Advanced Security, Partitioning

6. SQL Server

Editions: Express, Web, Standard, Enterprise
Windows Authentication: Domain integration
Always On: High availability

💾 Storage Types

General Purpose SSD (gp2/gp3)

Use case: Balanced price-performance
IOPS: 3 IOPS per GB (gp2), up to 16,000 (gp3)
Throughput: Up to 1,000 MiB/s (gp3)
Burst capability: Up to 3,000 IOPS for gp2

Provisioned IOPS (io1/io2)

Use case: I/O intensive workloads
IOPS: Up to 80,000 IOPS
Durability: 99.999% (io2)
IOPS:Storage ratio: 50:1 maximum

Magnetic (Standard)

Legacy storage: Không recommend cho production
Use case: Infrequent access
Performance: ~100 IOPS average

🔧 Instance Types

Burstable Performance (T3/T4g)

Use case: Variable workloads
CPU credits: Burst capability
Cost-effective: Development/testing

Memory Optimized (R5/R6i/X1e)

Use case: In-memory databases
High memory-to-vCPU ratio
SAP HANA workloads

Compute Optimized (C5/C6i)

Use case: CPU-intensive
High-frequency processors
Web servers, scientific computing

🌐 High Availability & Disaster Recovery

Multi-AZ Deployments

Primary AZ          Standby AZ
┌─────────────┐    ┌─────────────┐
│   Primary   │────│   Standby   │
│   Instance  │    │   Instance  │
│      ↓      │    │             │
│   EBS Vol   │────│   EBS Vol   │
└─────────────┘    └─────────────┘

Features: - Synchronous replication: No data loss - Automatic failover: 1-2 minutes - Single endpoint: Application transparency - Backup from standby: No I/O suspension

Read Replicas

Primary Instance
       ↓ (Async replication)
┌─────────────┐  ┌─────────────┐  ┌─────────────┐
│ Read Replica │  │ Read Replica │  │ Read Replica │
│   (AZ-a)     │  │   (AZ-b)     │  │ (Different  │
│              │  │              │  │  Region)    │
└─────────────┘  └─────────────┘  └─────────────┘

Features: - Read scaling: Distribute read traffic - Cross-region: Disaster recovery - Eventual consistency: Async replication - Promote to primary: Manual process

📊 Backup & Recovery

Automated Backups

Point-in-time recovery: 1-35 days retention
Daily snapshots: During backup window
Transaction logs: Continuous backup
No performance impact: Uses Multi-AZ standby

Manual Snapshots

User-initiated: On-demand backup
Retention: Until manually deleted
Cross-region copy: Disaster recovery
Share snapshots: Between accounts

Backup Best Practices

Backup Strategy:
  Automated Backups:
    Retention: 7 days (minimum for production)
    Window: Low-traffic period (e.g., 3:00-4:00 AM)

  Manual Snapshots:
    Before major changes: Application deployments
    Monthly: Long-term retention
    Cross-region: Disaster recovery

🔒 Security Features

Encryption

At rest: AES-256 encryption
In transit: SSL/TLS
Key management: AWS KMS
Encrypted snapshots: Automatic

Network Security

VPC: Private networking
Security Groups: Database-level firewall
Subnet Groups: Multi-AZ placement
VPC endpoints: Private connectivity

Access Control

IAM database authentication: Token-based
Master user: Database admin
Parameter groups: Configuration management
Option groups: Feature enablement

📈 Performance Optimization

Performance Insights

Query analysis: Top SQL statements
Wait events: Bottleneck identification
Historical data: 7 days free, extended with fee
Database load: Normalized metrics

Parameter Groups

-- Example MySQL parameters
innodb_buffer_pool_size = 70% of available memory
innodb_flush_log_at_trx_commit = 2  -- Better performance
slow_query_log = 1                  -- Enable slow query logging
long_query_time = 1                 -- Log queries > 1 second

Connection Pooling

# Application-level pooling
from sqlalchemy import create_engine

engine = create_engine(
    'mysql://user:pass@rds-endpoint:3306/db',
    pool_size=20,
    max_overflow=30,
    pool_recycle=3600
)

💰 Pricing Models

On-Demand Instances

Pay per hour: No commitment
Flexible: Start/stop anytime
Use case: Development, testing, unpredictable workloads

Reserved Instances

1 or 3 year terms: Up to 69% savings
Payment options: All upfront, partial, no upfront
Convertible: Change instance type

Aurora Serverless

Pay per second: When database is active
Auto-scaling: 0.5 to 256 ACUs
Auto-pause: Cost savings for intermittent workloads

🔄 Migration Strategies

Database Migration Service (DMS)

Source DB → DMS Replication Instance → Target RDS
   ↓              ↓                        ↓
On-premises → AWS DMS → AWS RDS
Oracle      → Continuous → PostgreSQL
             Replication   (with SCT)

Schema Conversion Tool (SCT)

Heterogeneous migrations: Oracle to PostgreSQL
Assessment report: Conversion complexity
Code conversion: Stored procedures, functions

Native Tools

# MySQL dump and restore
mysqldump -h source-host -u user -p database > backup.sql
mysql -h rds-endpoint -u user -p new-database < backup.sql

# PostgreSQL pg_dump
pg_dump -h source-host -U user -d database > backup.sql
psql -h rds-endpoint -U user -d new-database < backup.sql

🧪 Monitoring & Troubleshooting

CloudWatch Metrics

CPU Utilization: Instance performance
Database Connections: Connection count
Free Storage Space: Disk usage
IOPS: I/O performance
Latency: Read/Write response time

Enhanced Monitoring

OS-level metrics: Every 1-60 seconds
Process list: Running processes
Memory usage: Detailed memory metrics

Common Issues & Solutions

High CPU:
  Causes: Inefficient queries, missing indexes
  Solutions: Query optimization, read replicas

Storage Full:
  Causes: Rapid data growth, large transactions
  Solutions: Storage auto-scaling, log rotation

Connection Limits:
  Causes: Too many connections
  Solutions: Connection pooling, parameter tuning

Slow Queries:
  Causes: Missing indexes, lock contention
  Solutions: Query optimization, index creation

🚀 Best Practices

Performance Best Practices

Right-size instances: Monitor and adjust
Use read replicas: Scale read traffic
Optimize queries: Use EXPLAIN plans
Index strategy: Cover frequently queried columns
Connection pooling: Reduce connection overhead

Security Best Practices

Encryption: Enable at rest and in transit
VPC deployment: Private networking
Least privilege: Minimal database permissions
Regular patching: Keep engine updated
Backup verification: Test restore procedures

Cost Optimization

Reserved instances: For predictable workloads
Right-sizing: Match instance to workload
Storage optimization: Choose appropriate type
Delete unused: Remove old snapshots
Aurora Serverless: For intermittent workloads

📝 Exam Tips cho AWS SAA

Scenario-based Questions

High availability: Multi-AZ cho production
Read scaling: Read replicas cho heavy read workloads
Cross-region DR: Cross-region read replicas
Cost optimization: Reserved instances cho steady workloads

Key Differences

Multi-AZ vs Read Replicas:
  Multi-AZ: HA, automatic failover, same region
  Read Replicas: Read scaling, manual promotion, cross-region

Aurora vs Standard RDS:
  Aurora: Cloud-native, better performance, auto-scaling storage
  Standard: Traditional engines, predictable costs, established tools

📖 Tóm tắt

Amazon RDS cung cấp managed relational database service với: - Multiple database engines phù hợp với nhiều use cases - High availability với Multi-AZ deployments - Read scaling với read replicas - Comprehensive security features - Performance monitoring và optimization tools - Flexible pricing models cho different workloads

Hiểu rõ RDS deployment patterns, backup strategies, và performance optimization là critical cho AWS SAA exam.