
Splitting a Monolithic Database

#131 - Data Essentials in Software Architecture - Pramod Sadalage
Topic 7 of 8
Published May 1, 2023 · Clip 47:03–54:58 (7:56)
Pramod shares the five-step process from Software Architecture: The Hard Parts for splitting a monolithic database: (1) identify domain boundaries and which service owns which tables; (2) move owned tables into separate schemas within the same database, using synonyms for transparency; (3) give each service its own connection to its schema; then replicate schemas to separate machines and switch connections over, an iterative deploy-every-step approach for easy rollback. He covers edge cases where one service reads another's table, requiring APIs rather than cross-database connections, and advocates the strangler approach: extract one piece of functionality at a time, driven by business value and quality attributes like scalability and deployability, and stop splitting once goals are achieved.