Cloud Databases Explained: RDS, DynamoDB, BigQuery and More (2026)
Last verified: June 29, 2026 · Format: Breakdown · TechJacks Cloud Infrastructure Team: AWS Certified Solutions Architect + Google Cloud Professional Data Engineer
What Are Cloud Databases?
Cloud databases are managed database services hosted by a cloud provider. The provider handles infrastructure provisioning, patching, backups, and (in serverless offerings) scaling. You supply the schema and data; the provider operates the engine.
Govern the data behind your AI. The AI Data Governance & Quality Assessment: a checklist to keep your data trustworthy.
Your purchase helps keep our hubs free to read.
The core split is between two workload types:
| Type | Workload | Query pattern | Cloud examples |
|---|---|---|---|
| OLTP (Online Transaction Processing) | Live transactional data: orders, sessions, financial records | Fast row-level reads/writes, low latency (ms), high concurrency | Amazon RDS, Amazon DynamoDB, Azure SQL Database, Azure Cosmos DB, Google Cloud SQL, Google Firestore |
| OLAP (Online Analytical Processing) | Historical analysis, reporting, business intelligence | Large scans across many rows and columns, complex aggregations | Amazon Redshift, Azure Synapse Analytics, Google BigQuery |
Most cloud architectures use both: an OLTP database captures transactional data in real time, and a data warehouse receives periodic extracts for analytical workloads. Amazon Aurora and Amazon RDS for MySQL support zero-ETL integrations directly into Amazon Redshift (eliminating custom extract pipelines); DynamoDB also supports zero-ETL into Redshift for key-value workloads.
Relational Databases
Relational databases organize data into tables with defined schemas, enforce ACID transactions (Atomicity, Consistency, Isolation, Durability), and use SQL as the query language. They are the right tool for structured data with well-defined relationships: e-commerce inventory, financial ledgers, CRM records.
| Provider | Service | Engines supported | Managed features |
|---|---|---|---|
| AWS | Amazon RDS | PostgreSQL, MySQL, SQL Server, Oracle, MariaDB | Automated backups, Multi-AZ failover, read replicas, automated patching |
| AWS | Amazon Aurora | PostgreSQL-compatible, MySQL-compatible | Up to 5x MySQL throughput (per AWS), up to 15 read replicas, global databases, zero-ETL to Redshift |
| Azure | Azure SQL Database | SQL Server (PaaS fully managed) | Built-in HA, geo-replication, automatic tuning, elastic pools for multi-tenant apps |
| GCP | Cloud SQL | MySQL, PostgreSQL, SQL Server | Automated backups, point-in-time recovery, Private Service Connect, HA configuration |
| GCP | Cloud Spanner | GoogleSQL (PostgreSQL-compatible interface) | Horizontally scalable ACID, 99.999% SLA, global consistency — NewSQL for workloads that exceed single-region scale. Cloud SQL handles single-region managed MySQL/PostgreSQL/SQL Server; Spanner handles globally distributed write workloads. |
Shared responsibility: the cloud provider manages database engine infrastructure; you manage the schema, access control, query optimization, and application-level data handling. Customers own the data in all service models.
NoSQL Databases
NoSQL databases skip the fixed schema entirely. You get horizontal write scale and flexible data models; you give up JOIN semantics and, depending on the engine, some consistency guarantees. Choose NoSQL when your data structure evolves independently of a rigid schema, or when you need to scale writes across many nodes without resharding a relational table.
Why NoSQL consistency varies: CAP theorem. Distributed databases can guarantee at most two of three properties: Consistency (every read returns the latest write), Availability (every request gets a response), and Partition Tolerance (the system keeps working despite network splits). Since partition tolerance is non-negotiable in distributed cloud environments, NoSQL databases trade between Consistency and Availability — which is why DynamoDB offers configurable read consistency and Cosmos DB provides five consistency levels.
Consistency note: DynamoDB offers both eventually consistent reads (lower latency, lower cost per read) and strongly consistent reads. Cosmos DB exposes five configurable consistency levels from Eventual to Strong. If your workload requires that all nodes see the latest write immediately, confirm your chosen service supports strong consistency — not all NoSQL engines guarantee this by default.
| Model | Best for | AWS | Azure | GCP |
|---|---|---|---|---|
| Key-value | Session stores, shopping carts, caching | DynamoDB (also document), ElastiCache (Redis/Memcached) | Azure Cache for Redis, Cosmos DB (Table API) | Memorystore (Redis) |
| Document | User profiles, content management, catalogs | DynamoDB (JSON items), DocumentDB | Cosmos DB (NoSQL API) | Firestore, MongoDB Atlas on GCP |
| Wide-column | IoT telemetry, time-series at scale, high write throughput | Amazon Keyspaces (Cassandra-compatible) | Cosmos DB (Cassandra API), Azure Table Storage | Bigtable |
| In-memory | Sub-millisecond caching, rate limiting, leaderboards | ElastiCache (Redis, Memcached) | Azure Cache for Redis | Memorystore (Redis, Memcached) |
| Graph | Social networks, fraud detection, knowledge graphs, recommendation engines | Amazon Neptune | Cosmos DB (Gremlin API) | Neo4j AuraDB (GCP Marketplace) |
Amazon DynamoDB is a key-value/document database (not wide-column) classified as an OLTP service in the AWS taxonomy. It supports zero-ETL integrations directly into Amazon Aurora and Amazon RDS for MySQL, with Redshift as an analytical target. It scales horizontally across AWS regions and is priced on read/write capacity units (provisioned) or on-demand request pricing — see aws.amazon.com/dynamodb/pricing/ for current RCU/WCU rates.
Data Warehouses
Data warehouses store large volumes of historical data in columnar format and are built for complex aggregation queries, not individual row lookups.
| Service | Provider | Pricing model | Scaling | Idle charges |
|---|---|---|---|---|
| BigQuery | GCP | On-demand: $6.25/TiB scanned. Slots: $0.04/slot-hour (Standard) to $0.10/slot-hour (Enterprise Plus) | Fully serverless; no instance management | None (serverless) |
| Redshift Serverless | AWS | $1.50/hr minimum billing floor (per-second billing, 60s minimum; per AWS documentation). RPU range scales to query demand. Reserved pricing available — see aws.amazon.com/redshift/pricing/ for current Savings Plan rates. No charge when idle. | Scales RPUs automatically; auto-shuts down when idle | None (shuts down automatically) |
| Redshift Provisioned | AWS | Starts at $0.543/hr on-demand (RA3.large). Larger instance types available — see aws.amazon.com/redshift/pricing/ for full RA3 instance family and Reserved Instance rates. | Manual cluster resize or elastic resize | Yes (always-on cluster) |
| Synapse Dedicated SQL | Azure | Storage: $23/TB/month (1 TB blocks). Compute billed by DWU-hour at the HIGHEST scale used that hour. See azure.microsoft.com for current DWU rates. | Manual DWU scaling with pause/resume | Compute: No when paused (pausing stops compute billing). Storage: Yes — $23/TB/month continues even while paused, including 7-day snapshots. |
| Synapse Serverless SQL | Azure | $5/TB processed; no instance management | Fully serverless | None (serverless) |
Pricing Comparison: Data Warehouses
Warehouse pricing varies by compute model. BigQuery on-demand is cheapest for teams running analytical queries a few hours a day — you pay only for bytes scanned. Redshift Serverless removes idle risk entirely: it shuts down between queries, so you pay nothing when it's not running. Provisioned clusters (Redshift RA3, Synapse Dedicated) pay off only at sustained, high-volume workloads where you can commit to reserved pricing.
Key grounded rates (verified June 29, 2026; always confirm on provider pricing pages before budgeting):
| Metric | BigQuery | Redshift Serverless | Synapse Dedicated (DW100c) |
|---|---|---|---|
| On-demand compute | $6.25/TiB scanned | $1.50/hr minimum (4 RPU floor, per-second billing, 60s minimum; no charge when idle). Per-RPU rate: see aws.amazon.com/redshift/pricing/ | Compute billed per DWU-hour at the highest scale used that hour. See azure.microsoft.com/pricing for current DWU rates |
| Minimum charge | 1 TiB/month free tier | $1.50/hr billing floor per active session | Hourly billing always active |
| Storage | Active: $0.02/GB/month; Long-term (90-day unmodified): ~50% discount | Separate S3 storage | $23/TB/month (includes 7-day incremental snapshots) |
| Idle billing | None | None (auto-shutdown) | Compute billed per DWU-hour even when idle. Pausing stops compute billing; storage ($23/TB/month) continues. Any active query in an hour bills that full hour at the highest DWU tier used. |
Rates change without notice. All figures verified from official provider pricing pages June 29, 2026. Cloud providers update pricing without advance announcement — always confirm current rates before budgeting or committing to reserved capacity. See AWS Redshift, BigQuery, and Synapse Analytics pricing pages for current rates.
Cost Gotcha Register
These billing behaviors are verified from official provider pricing documentation (June 2026). Each one has an easy-to-miss trigger that generates unexpected charges.
LIMIT 10 clause at the end of a 1 TiB query still bills for 1 TiB. Use table partitioning and clustering to reduce actual bytes scanned.
Prevention: Always run SELECT * EXCEPT (large_columns) or scope your column list before testing on large tables. Use BigQuery's built-in cost estimator (--dry_run in bq CLI or "This query will process X GB" in the Console) before executing.
INSERT, UPDATE, or DELETE) resets the 90-day timer for that table. Batch-load tables by partition where possible to preserve the discount on older partitions.Database Service Picker
Test Your Knowledge
Amazon RDS, Amazon Redshift, Amazon DynamoDB, and Amazon Aurora are trademarks of Amazon Web Services, Inc. Microsoft Azure SQL Database, Azure Synapse Analytics, and Azure Cosmos DB are trademarks of Microsoft Corporation. Google Cloud SQL, Google BigQuery, and Google Firestore are trademarks of Google LLC. All pricing figures are sourced from official provider pricing pages as of June 2026 and may change. Verify current rates at aws.amazon.com/pricing, azure.microsoft.com/pricing, and cloud.google.com/pricing before budgeting or committing resources.