SQL vs NoSQL Decision Framework

During system design interviews and architecture RFCs, evaluate these 8 architectural axes before selecting a database engine:

1. Transactional Guarantee: Do you require strict ACID multi-row atomicity and foreign key referential integrity (e.g., banking, e-commerce checkout), or is eventual consistency (BASE) acceptable?
2. Data Model Structure: Is data relational and tabular, hierarchical/document (JSON), graph/networked, or simple key-value pairs?
3. Read vs Write Ratio: Is the system read-heavy (100:1 → Caching / Read Replicas) or write-heavy (1:10 → LSM-Tree Cassandra/Kafka)?
4. Query Complexity: Do you need complex ad-hoc queries, aggregations (GROUP BY), and multi-table joins, or strictly primary-key lookups?
5. Schema Stability: Is the schema well-defined and stable (SQL DDL), or rapidly evolving and polymorphic (JSON document)?
6. Scale & Storage Volume: Will total data exceed 5 Terabytes, and write QPS exceed 50,000 writes/second (requiring horizontal sharding)?
7. Latency SLAs: Do queries require sub-millisecond memory speed (<1ms → Redis), or is 5–20ms acceptable (Disk RDBMS)?
8. Operational Maturity: Does your engineering team have the operational expertise to maintain distributed Cassandra/Kafka clusters?