本文由 AI 分析生成
建立時間： 2026-03-24 來源： https://cloud.google.com/architecture/database-migration-concepts-principles-part-1
Summary
Google Cloud Architecture Center’s comprehensive guide to database migration concepts: covering terminology, homogeneous vs. heterogeneous migration, downtime tolerance tradeoffs, CDC vs. differential querying, data transformation patterns, and common pitfalls (order violations, consistency issues, missing/duplicate data).
Google Cloud 架構中心對資料庫遷移概念的全面指南，涵蓋術語定義、同質與異質遷移、停機容忍度取捨、CDC 與差異查詢方法、數據轉換模式以及常見陷阱。
Key Points
- Homogeneous migration: same DBMS (e.g., MySQL → MySQL); schemas often identical, can use native replication
- Heterogeneous migration: different DBMS (e.g., Oracle → Spanner); requires schema/data transformation
- CDC (Change Data Capture) via transaction logs is preferred over differential querying for minimal load impact
- Downtime spectrum: zero (near-zero) → minimal → significant; each enables simpler migration tooling
- Migration cardinality: 1:1 (direct), n:1 (consolidation), 1:n (distribution), n:m (redistribution)
- Key pitfalls: order violations in parallel migration, consistency violations from differential query timestamps, flashback after source DB failover
Insights
The “active-active migration” concept (both source and target accept writes simultaneously) is underappreciated but complex — it requires conflict resolution rules and is rarely the right choice. The insight that migration consistency (complete, duplicate-free, ordered) is equivalent to saying “source and target state must be equivalent at migration end” cleanly captures what correctness means without prescribing implementation. The CDC-based approach through transaction logs (binary log in MySQL) is almost always preferable to polling because it’s already built into the database and imposes minimal additional load.
Connections
- change data capture
- database replication
- Google Cloud Migration
- PostgreSQL logical replication
- ETL vs CDC
Raw Excerpt
CDC is the preferred approach for capturing changes in a database management system. CDC is built into the database itself and has the least load impact on the system.