Workload-Specific Database Design - AWS Database Specialty¶

Overview¶

Workload-specific database design covers 26% of the DBS-C01 exam - the largest domain. This domain focuses on choosing the right database service for specific use cases and understanding the design considerations for each.

📖 AWS Database Services - Overview of all database offerings

Amazon RDS¶

📖 RDS User Guide - Relational Database Service

Supported Engines¶

Key Features¶

• Multi-AZ: Synchronous standby replica, automatic failover (< 60 sec)
• Read Replicas: Asynchronous replication, can be cross-region
• Storage Types: gp3 (general), io1/io2 (provisioned IOPS), magnetic (legacy)
• Automated Backups: Point-in-time recovery, 0-35 day retention
• Manual Snapshots: User-initiated, unlimited retention
• RDS Proxy: Connection pooling, reduces failover time by 66%

📖 Multi-AZ Deployments - High availability 📖 Read Replicas - Read scaling 📖 RDS Proxy - Connection management

When to Use RDS¶

• Traditional relational workloads (OLTP)
• Existing applications using MySQL, PostgreSQL, Oracle, or SQL Server
• Need for complex queries, joins, and ACID transactions
• Do not need more than 64 TB storage or microsecond latency

Amazon Aurora¶

📖 Aurora User Guide - Cloud-native relational database

Architecture¶

• Storage layer: 6 copies across 3 AZs, auto-scales 10 GB to 128 TB
• Quorum writes (4/6) and reads (3/6) for high durability
• Continuous backup to S3 with 1-second granularity
• Storage is independent of compute - shared across replicas
• Redo log-based replication (not binlog) - <100ms replica lag

Aurora Variants¶

Aurora Provisioned: - Writer instance + up to 15 read replicas - Automatic failover to replica in < 30 seconds - Reader endpoint for load-balanced reads

Aurora Serverless v2: - Scales between 0.5 and 128 ACUs (Aurora Capacity Units) - Fine-grained scaling in 0.5 ACU increments - Sub-second scaling response - Good for variable or unpredictable workloads

Aurora Global Database: - Cross-region replication with < 1 second lag - Up to 5 secondary regions - RPO < 1 second, RTO < 1 minute for planned failover - Write forwarding from secondary to primary region - Managed planned failover and unplanned (detach) failover

📖 Aurora Global Database - Cross-region replication 📖 Aurora Serverless v2 - Auto-scaling compute

When to Use Aurora¶

• Need better performance than standard RDS (5x MySQL, 3x PostgreSQL)
• Need more than 5 read replicas
• Need auto-scaling storage beyond 64 TB
• Need cross-region disaster recovery with low RPO
• Variable workloads (Serverless v2)

Amazon DynamoDB¶

📖 DynamoDB Developer Guide - NoSQL key-value and document

Core Concepts¶

• Tables, Items, Attributes: Schema-less except for primary key
• Primary Key: Partition key (hash) or partition key + sort key (hash + range)
• Secondary Indexes: GSI (any attribute as key, separate throughput) and LSI (same partition key, different sort key, created at table creation)
• Capacity Modes: On-demand (pay per request) or Provisioned (specify RCU/WCU)

Key Features¶

• Single-digit millisecond latency at any scale
• Automatic multi-AZ replication (3 AZs)
• DAX (DynamoDB Accelerator): In-memory cache, microsecond reads
• Global Tables: Multi-region, multi-active replication
• Streams: Ordered change data capture (24-hour retention)
• TTL: Automatic item expiration
• PartiQL: SQL-compatible query language for DynamoDB

📖 DynamoDB Best Practices - Design patterns 📖 Global Tables - Multi-region

Data Modeling¶

• Design for access patterns first (not entities)
• Single table design: Store multiple entity types in one table
• Composite sort keys: Combine attributes for flexible querying
• Sparse indexes: GSI on attribute that not all items have
• Overloaded GSI: Reuse GSI for multiple query patterns
• Avoid hot partitions: Distribute writes evenly across partition keys

When to Use DynamoDB¶

• Key-value or simple document access patterns
• Need single-digit millisecond latency
• Need automatic scaling with unpredictable traffic
• Session management, shopping carts, gaming leaderboards
• IoT data with high write throughput

Amazon DocumentDB¶

📖 DocumentDB Developer Guide - MongoDB-compatible document database

• MongoDB 3.6, 4.0, 5.0 API compatibility
• Storage: Up to 128 TB, 6 copies across 3 AZs (Aurora-like architecture)
• Up to 15 read replicas
• Automatic failover, continuous backup to S3
• Not all MongoDB features supported (check compatibility matrix)

When to Use DocumentDB¶

• MongoDB workloads on AWS with managed infrastructure
• Document-oriented data with flexible schema
• Need managed scaling and high availability
• Migrating from self-managed MongoDB

Amazon Neptune¶

📖 Neptune User Guide - Graph database

• Supports Property Graph (Gremlin/openCypher) and RDF (SPARQL)
• Up to 15 read replicas, 6 copies across 3 AZs
• Neptune Serverless: Auto-scaling capacity
• Neptune ML: Graph neural networks via SageMaker integration
• Neptune Analytics: Analytical queries on large graphs

When to Use Neptune¶

• Social networks (friends-of-friends queries)
• Fraud detection (relationship patterns)
• Knowledge graphs, recommendation engines
• Identity graphs, network topology
• Any workload where relationships between data are primary

Amazon ElastiCache¶

📖 ElastiCache User Guide - In-memory caching

Redis vs Memcached¶

When to Use ElastiCache¶

• Database query result caching
• Session store management
• Real-time leaderboards (Redis sorted sets)
• Pub/sub messaging (Redis)
• Simple caching with multi-threading (Memcached)

Amazon Keyspaces¶

📖 Keyspaces Developer Guide - Apache Cassandra compatible

• Serverless, Cassandra Query Language (CQL) compatible
• On-demand or provisioned throughput
• Encryption at rest, point-in-time recovery
• Single-digit millisecond latency

When to Use Keyspaces¶

• Migrating Cassandra workloads to managed service
• Wide-column time-series data
• High-volume write workloads with simple read patterns

Amazon QLDB¶

📖 QLDB Developer Guide - Quantum Ledger Database

• Immutable, cryptographically verifiable transaction log
• Append-only journal with SHA-256 hash chain
• PartiQL query language
• Serverless with automatic scaling
• Full audit history of all changes

When to Use QLDB¶

• Financial transactions requiring audit trail
• Supply chain tracking with verifiable history
• Regulatory compliance requiring immutable records
• Any application needing cryptographic proof of data integrity

Amazon Timestream¶

📖 Timestream Developer Guide - Time-series database

• Purpose-built for time-series data
• Automatic tiering: in-memory (recent) to magnetic (historical)
• Built-in time-series functions (interpolation, smoothing, approximation)
• 1000x faster and 1/10^th cost of relational for time-series queries
• SQL query interface with time-series extensions

When to Use Timestream¶

• IoT sensor data and telemetry
• Application and infrastructure monitoring metrics
• DevOps monitoring and alerting
• Industrial equipment monitoring

Database Selection Decision Tree¶

Common Exam Patterns¶

1. "Microsecond reads on DynamoDB" - DAX (not ElastiCache)
2. "Multi-region active-active relational" - Not natively supported; use Aurora Global Database (active-passive with write forwarding)
3. "Multi-region active-active NoSQL" - DynamoDB Global Tables
4. "Audit trail with cryptographic verification" - QLDB
5. "Graph relationships" - Neptune
6. "MongoDB migration" - DocumentDB
7. "Cassandra migration" - Keyspaces
8. "Variable relational workload" - Aurora Serverless v2
9. "Cross-region disaster recovery < 1 sec RPO" - Aurora Global Database
10. "Connection pooling for Lambda" - RDS Proxy