AI Glossary / Data engineering and architecture / Beginner
Cloud database
Definition
A cloud database is a database deployed using cloud computing resources. It may be operated by the customer on cloud infrastructure or delivered as a managed database service. It stores and retrieves data through a database engine while relying on cloud resources for some combination of computing, storage, networking and operational management.
What it means
The term describes deployment and service delivery, not one data model. A cloud database can be relational, document-oriented, key-value, graph-based or another type. A managed service may handle backups, patching or failover, but the exact division of responsibility varies. The customer still needs appropriate data design, access rules and application behavior. Moving an existing database to a cloud virtual machine does not automatically give it every capability of a managed or distributed cloud database.
How it works
Applications connect through defined endpoints and authenticate with permitted identities. The engine processes queries and updates according to its model and transaction rules. Cloud resources provide the underlying capacity, and management services may automate selected operating tasks. Replication, backup, scaling and recovery depend on the product and configuration. The application must understand consistency, connection limits, timeouts and failover behavior. A service can hide hardware management while still requiring careful schema, index and query design.
Three practical examples
- A small business uses a managed relational database for orders and customer accounts while controlling application access and data structure.
- A mobile application stores flexible document records in a cloud document database, using the service's supported query and consistency rules.
- An AI application stores metadata and permissions for retrieved documents in a cloud database, ensuring that retrieval logic respects the current user.
Think of it this way
Think of renting a managed records facility. The facility can supply space, maintenance and retrieval equipment, but the customer must still organize records sensibly and decide who may access them. The analogy is limited because a database also executes queries and transactions, and different services assume different amounts of operating responsibility.
Worked example
Suppose an application makes 200 requests per second and each request issues five database queries. That produces 1,000 queries per second before background jobs and retries. If a redesign combines two related lookups and reduces the total to three queries per request, the application produces 600 queries per second, a 40% reduction. This does not prove a 40% cost or latency reduction because query complexity, caching, billing and contention differ. It illustrates that application design can change database demand before any cloud scaling decision is made.
Illustrative case study
Fictional case: Cedar Booking adopts a managed cloud database and assumes automatic backups mean recovery is fully solved. A rehearsal reveals that the team has never defined which point in time it needs to restore or how the application will reconnect afterward. It documents recovery steps, tests them with nonproduction data and reviews access permissions. This fictional case illustrates responsibilities that remain despite managed infrastructure; it does not describe a real outage.
Common misconceptions and mistakes
Do not equate replication with backup: replicas can repeat an unwanted deletion or corruption event. Do not assume scaling is instantaneous, unlimited or free of application impact. A relational cloud database does not become NoSQL merely because it is hosted remotely. Also, a managed label does not guarantee that every setting is appropriate for the workload. Query plans, indexes, connection pools and data retention remain important even when the provider handles servers.
Business use
Businesses should choose a cloud database based on required queries, transactions, availability, recovery and operating skills. Include migration, data transfer, exit options and ongoing usage costs. Test representative workloads and failure behavior rather than comparing only maximum advertised capacity. For AI systems, consider vector or retrieval features as part of the whole data model, including permissions, updates and provenance, rather than treating one search capability as the entire database requirement.
People also ask
Is every cloud database managed?
No. A customer can run a database on cloud infrastructure and manage it directly.
Does cloud hosting determine whether the database uses SQL?
No. Deployment and data model are separate choices.
Can applications lose access during failover?
They can experience interruption or changed connections, depending on the service and configuration, so client behavior should be tested.
Check your understanding
A cloud database has a read replica containing the same recent data as the primary. Does that replica necessarily recover a table accidentally deleted yesterday?
Answer: No. The deletion may already have replicated. Recovery requires a suitable backup or point-in-time recovery capability configured and tested for that need.
Related terms
- Relational database
- Replication (statistics)
- Backup
- Point-in-time recovery
- database as a service
- NoSQL
- replication
- connection pool
Learn next
List the queries and updates one application needs, then map them to a candidate database's documented guarantees. Run a small workload and a recovery rehearsal. Record what the provider manages, what the team configures and what the application must do after a transient failure.
More in Data engineering and architecture
- Client-Server Architecture - Client-server architecture organizes software into components that request services and components that provide them.
- Client-side persistent data - Client-side persistent data are information saved on a user's device so that an application can use them beyond the current in-memory session.
- Cloud analytics - Cloud analytics is the use of cloud-based computing and data services to examine data and produce insights, reports or analytical models.
- Cloud-based integration - Cloud-based integration connects applications, data and processes using capabilities hosted in a cloud environment.
- Cloud computing architecture - Cloud computing architecture is the arrangement of components, responsibilities and connections that delivers a cloud-based system.
- Cloud computing security - Cloud computing security is the protection of cloud-hosted systems, data and services against unauthorized access, misuse, disruption and unintended exposure.
Sources
Primary provider references: AWS cloud database explanation, https://aws.amazon.com/what-is/cloud-database/ ; Amazon RDS getting-started documentation, https://docs.aws.amazon.com/AmazonRDS/latest/gettingstartedguide/rds-gsg.pdf . Official provider search excerpts consulted; full product-specific and editorial review remain pending. The entry does not assume identical capabilities across cloud databases.
Previous: Cloud computing security Next: Cloud-native computing