Cloud Database Services Cheat Sheet

Updated 2026-05-25

Next Topic: Cloud Deployment Archetypes Cheat Sheet

🧠Study flashcards on this topic107 cards · spaced repetition→

Cloud database services represent fully managed or serverless database platforms provided by cloud vendors like AWS, Azure, and Google Cloud, eliminating the operational burden of provisioning, patching, and scaling infrastructure. These services enable organizations to focus on application development rather than database administration, offering built-in features like automated backups, high availability, and global distribution. In 2026, the landscape has expanded beyond traditional relational and NoSQL models to include purpose-built vector databases for AI workloads, distributed NewSQL systems offering global ACID transactions, and HTAP databases unifying transactional and analytical processing in a single cluster. A key design consideration: managed services abstract complexity but may limit low-level control—choosing the right service requires balancing flexibility, cost, and operational simplicity for your specific workload characteristics.

What This Cheat Sheet Covers

This topic spans 13 focused tables and 113 indexed concepts, 107 flashcards. Below is a complete table-by-table outline of this topic, spanning foundational concepts through advanced details.

Table 1: Database Service ModelsTable 2: Major Cloud Database EnginesTable 3: High Availability ConfigurationsTable 4: Scaling StrategiesTable 5: Backup and Recovery StrategiesTable 6: Performance OptimizationTable 7: Security and Access ControlTable 8: Multi-Region DeploymentTable 9: Database Migration StrategiesTable 10: Monitoring and ObservabilityTable 11: Storage and Cost OptimizationTable 12: Consistency Models and Trade-offsTable 13: Vector Database and AI Search Services

Quick IndexSubscribe to unlock

A jump-to index of every table row in this cheat sheet.

Mind MapSubscribe to unlock

An interactive map of every table and concept in this topic.

Table 1: Database Service Models

Choosing the right database model is the single most consequential architectural decision in a cloud system. Matching your data structure and access patterns to the right model—relational, document, key-value, vector, or time-series—determines throughput, consistency, and cost at every scale.

Type	Example	Description
Relational (SQL)	`AWS RDS PostgreSQL` `Azure SQL Database`	• Fully managed SQL databases with ACID guarantees, schema enforcement, and support for complex joins • best for transactional systems, structured data, and applications requiring strong consistency.
NoSQL Document	`MongoDB Atlas` `DynamoDB` `Cosmos DB`	• Schema-flexible document stores optimized for hierarchical data and rapid development • ideal for content management, user profiles, and applications with evolving data models.
NoSQL Key-Value	`DynamoDB` `Azure Cosmos DB`	• Ultra-fast lookups by primary key with single-digit millisecond latency • designed for session stores, caching layers, and high-throughput read/write operations at massive scale.
In-Memory	`Amazon MemoryDB` `ElastiCache Serverless` `Azure Managed Redis`	• Entire dataset stored in RAM for microsecond reads and single-digit millisecond writes • durable in-memory databases (MemoryDB) add multi-AZ transaction logs; pure caches (ElastiCache) are for ephemeral acceleration.
Vector	`pgvector` `Pinecone` `OpenSearch k-NN`	• Store high-dimensional embeddings alongside structured data for semantic similarity search • essential for RAG pipelines, semantic search, and recommendation systems in AI applications; supported natively via pgvector in most cloud PostgreSQL services.
Columnar/Analytical	`BigQuery` `Snowflake` `Redshift`	• Column-oriented storage for analytical queries across billions of rows • designed for data warehousing, business intelligence, and OLAP workloads with aggregations over large datasets.
Time Series	`Amazon Timestream` `InfluxDB Cloud` `TimescaleDB`	• Optimized for timestamped metrics and events with automatic data retention policies and downsampling • critical for IoT telemetry, DevOps monitoring, and financial tick data.
NewSQL / Distributed SQL	`CockroachDB` `TiDB` `Google Cloud Spanner`	• Horizontally scalable relational databases delivering SQL semantics with automatic sharding and ACID transactions across nodes and regions • eliminates the manual sharding required by traditional RDBMS at global scale
Graph	`Amazon Neptune` `Neo4j Aura` `Cosmos DB Gremlin`	• Model relationships as first-class citizens using nodes and edges • excel at fraud detection, recommendation engines, social networks, and knowledge graphs where connections matter more than entities.
Wide-Column	`Bigtable` `Cassandra`	• Distributed storage for sparse, high-dimensional data with horizontal scalability • used for time-series at extreme scale, user activity logs, and IoT sensor networks.
Serverless	`Aurora Serverless v2` `Cosmos DB Serverless` `Firestore`	• Auto-scaling capacity with per-request billing and zero administration • eliminates capacity planning for intermittent workloads, development environments, and unpredictable traffic patterns.

Table 2: Major Cloud Database Engines

The three hyperscalers each offer purpose-built database services that go well beyond generic SQL or NoSQL—understanding the differences between Aurora, Cosmos DB, Spanner, AlloyDB, and their peers prevents costly migrations later. For each service, the key questions are: what engine does it run, what consistency model does it enforce, and at what scale does its pricing model become prohibitive?

Service	Example	Description
Amazon RDS	`rds:postgres:16` `rds:mysql:8.0`	• Managed relational database supporting six engines (PostgreSQL, MySQL, MariaDB, Oracle, SQL Server, Db2) • offers automated backups, patching, and Multi-AZ failover with minimal administrative overhead.

The rest of this section is part of the full cheat sheet

Table 3: High Availability Configurations

Cloud database HA is not a simple on/off switch — it is a spectrum from same-zone standby replicas to global active-active clusters. The choice between synchronous Multi-AZ, asynchronous read replicas, and global tables determines your actual RPO and RTO under failure, not just your SLA number.

Pattern	Example	Description
Multi-AZ Deployment	`Primary: us-east-1a` `Standby: us-east-1b`	• Synchronous replication to a standby instance in a different availability zone with automatic failover typically completing in 60–120 seconds • protects against AZ failures with zero data loss (RPO=0).

The rest of this section is part of the full cheat sheet

Table 4: Scaling Strategies

Databases scale in fundamentally different ways depending on whether the bottleneck is reads, writes, connections, or storage. Combining vertical scaling, read replicas, auto-scaling, and connection proxies covers most production workloads; sharding is a last resort that trades simplicity for unlimited horizontal scale.

Technique	Example	Description
Vertical Scaling (Scale Up)	`db.t3.medium → db.r6g.2xlarge`	• Increase CPU, memory, and IOPS by upgrading to a larger instance type • simple to implement but limited by maximum instance size and typically requires brief downtime (1–5 minutes for modern managed services).

The rest of this section is part of the full cheat sheet

Table 1: Database Service Models

Type	Example	Description
Relational (SQL)	`AWS RDS PostgreSQL` `Azure SQL Database`	• Fully managed SQL databases with ACID guarantees, schema enforcement, and support for complex joins • best for transactional systems, structured data, and applications requiring strong consistency.
NoSQL Document	`MongoDB Atlas` `DynamoDB` `Cosmos DB`	• Schema-flexible document stores optimized for hierarchical data and rapid development • ideal for content management, user profiles, and applications with evolving data models.
NoSQL Key-Value	`DynamoDB` `Azure Cosmos DB`	• Ultra-fast lookups by primary key with single-digit millisecond latency • designed for session stores, caching layers, and high-throughput read/write operations at massive scale.
In-Memory	`Amazon MemoryDB` `ElastiCache Serverless` `Azure Managed Redis`	• Entire dataset stored in RAM for microsecond reads and single-digit millisecond writes • durable in-memory databases (MemoryDB) add multi-AZ transaction logs; pure caches (ElastiCache) are for ephemeral acceleration.
Vector	`pgvector` `Pinecone` `OpenSearch k-NN`	• Store high-dimensional embeddings alongside structured data for semantic similarity search • essential for RAG pipelines, semantic search, and recommendation systems in AI applications; supported natively via pgvector in most cloud PostgreSQL services.
Columnar/Analytical	`BigQuery` `Snowflake` `Redshift`	• Column-oriented storage for analytical queries across billions of rows • designed for data warehousing, business intelligence, and OLAP workloads with aggregations over large datasets.
Time Series	`Amazon Timestream` `InfluxDB Cloud` `TimescaleDB`	• Optimized for timestamped metrics and events with automatic data retention policies and downsampling • critical for IoT telemetry, DevOps monitoring, and financial tick data.
NewSQL / Distributed SQL	`CockroachDB` `TiDB` `Google Cloud Spanner`	• Horizontally scalable relational databases delivering SQL semantics with automatic sharding and ACID transactions across nodes and regions • eliminates the manual sharding required by traditional RDBMS at global scale
Graph	`Amazon Neptune` `Neo4j Aura` `Cosmos DB Gremlin`	• Model relationships as first-class citizens using nodes and edges • excel at fraud detection, recommendation engines, social networks, and knowledge graphs where connections matter more than entities.
Wide-Column	`Bigtable` `Cassandra`	• Distributed storage for sparse, high-dimensional data with horizontal scalability • used for time-series at extreme scale, user activity logs, and IoT sensor networks.
Serverless	`Aurora Serverless v2` `Cosmos DB Serverless` `Firestore`	• Auto-scaling capacity with per-request billing and zero administration • eliminates capacity planning for intermittent workloads, development environments, and unpredictable traffic patterns.

Table 2: Major Cloud Database Engines

Service	Example	Description
Amazon RDS	`rds:postgres:16` `rds:mysql:8.0`	• Managed relational database supporting six engines (PostgreSQL, MySQL, MariaDB, Oracle, SQL Server, Db2) • offers automated backups, patching, and Multi-AZ failover with minimal administrative overhead.

Table 3: High Availability Configurations

Pattern	Example	Description
Multi-AZ Deployment	`Primary: us-east-1a` `Standby: us-east-1b`	• Synchronous replication to a standby instance in a different availability zone with automatic failover typically completing in 60–120 seconds • protects against AZ failures with zero data loss (RPO=0).

Table 4: Scaling Strategies

Technique	Example	Description
Vertical Scaling (Scale Up)	`db.t3.medium → db.r6g.2xlarge`	• Increase CPU, memory, and IOPS by upgrading to a larger instance type • simple to implement but limited by maximum instance size and typically requires brief downtime (1–5 minutes for modern managed services).