Comparing managed database options by workload

Compare managed databases against the work they must do. A transactional application, a document-based catalogue and an analytical reporting system need different checks.

Shortlist a database type first. Then verify the features and limits of the exact service, edition, tier and region.

Set a common comparison

Record the required data model and engine, transaction rules, read and write mix, peak concurrency, data volume, recovery targets and application location. Include required drivers, extensions, parameters and privileges.

For viable candidates, test the same representative data and queries. Record the configuration used.

Key Limits and Features by Service

Amazon RDS: an existing relational application

RDS is a candidate when the application already uses one of its supported relational engines and the team wants a managed DB instance. It provides resizable capacity and manages common database administration tasks.

Check engine-version and region support, required features, storage performance and the restore route. The service documentation includes quotas and constraints, including the maximum number of database connections.

Familiar SQL tooling helps only if the proposed configuration meets the workload.

Cloud SQL for PostgreSQL: PostgreSQL on Google Cloud

Cloud SQL may suit an application that uses PostgreSQL and operates on Google Cloud. It restricts some superuser operations and extensions compared with a self-hosted installation.

Enterprise and Enterprise Plus editions differ in point-in-time log retention, and managed connection pooling is listed for Enterprise Plus. Verify the required extension, edition, recovery settings and connection plan on the proposed instance.

Azure Database for PostgreSQL flexible server: PostgreSQL on Azure

Azure's flexible server offers Burstable, General Purpose and Memory Optimized tiers. Maximum user connections vary by tier and vCores, with some connections reserved for service operations. Storage cannot be reduced on an existing server.

Microsoft describes Burstable as primarily suited to non-production use because sustained load can exhaust CPU credits. Check capacity, recovery and storage growth on the intended tier rather than treating tier names as a performance result.

MongoDB Atlas: document-shaped application data

Atlas is a candidate when records and queries suit MongoDB's document model. Test the schema, indexes and transaction requirements.

Its Flex tier is positioned for learning MongoDB or developing small proof-of-concept applications; it is not a substitute for checking a production tier.

Flex lacks continuous backup with point-in-time restore and private endpoints, and it has documented storage limits. Check the required capabilities and region on the chosen Atlas deployment type.

BigQuery: analytical queries

BigQuery is a managed analytical platform to consider for reporting and aggregation. Assess data-loading work, query frequency, reporting freshness and bytes processed. Evaluate the live application's transaction store separately.

Decide from evidence

For each candidate, mark data correctness, required features, peak latency, recovery, connection headroom, location and estimated cost as pass, fail or unknown. An unknown calls for verification.