Shared-database-per-service pattern - AWS Prescriptive Guidance

Shared-database-per-service pattern

In the shared-database-per-service pattern, the same database is shared by several microservices. You need to carefully assess the application architecture before adopting this pattern, and make sure that you avoid hot tables (single tables that are shared among multiple microservices). All your database changes must also be backward-compatible; for example, developers can drop columns or tables only if objects are not referenced by the current and previous versions of all microservices.

In the following illustration, an insurance database is shared by all the microservices and an IAM policy provides access to the database. This creates development time coupling; for example, a change in the "Sales" microservice needs to coordinate schema changes with the "Customer" microservice. This pattern does not reduce dependencies between development teams, and introduces runtime coupling because all microservices share the same database. For example, long-running "Sales" transactions can lock the "Customer" table and this blocks the "Customer" transactions.

Shared-database-per-service pattern

You should consider using this pattern if:

  • You don't want too much refactoring of your existing code base.

  • You enforce data consistency by using transactions that provide atomicity, consistency, isolation, and durability (ACID).

  • You want to maintain and operate only one database.

  • Implementing the database-per-service pattern is difficult because of interdependencies among your existing microservices.

  • You don’t want to completely redesign your existing data layer.