Cassandra

Interview Trigger

看到这些关键词想到 Cassandra：

Huge Write Throughput / Time-series / Distributed
Eventually Consistent / Petabytes / Multi-region

What

A distributed wide-column NoSQL database optimized for high write throughput and horizontal scalability.

为什么不用 MySQL？

假设每天 10 Billion Events (IoT, Log, Telemetry, Sensor)。MySQL 写压力越来越大。Cassandra 天然 Horizontal Scaling，增加 Node 继续写。

缺点

没有复杂 JOIN，没有强事务。通常 Eventually Consistent。

Coinbase 场景

✅ User Activity Log / Price Tick / Metrics / Audit Log
❌ Wallet / Transfer / Payment

Use Cassandra when write scalability is more important than strong consistency.

MongoDB

Interview Trigger

Flexible Schema / JSON / Rapid Development / Nested Document

为什么不用 MySQL？

每个 Product 字段完全不同（Phone: cpu/ram/camera, Book: author/publisher），Schema 一直变化，Mongo 方便。

缺点

复杂 Transaction。关系查询没有 MySQL 成熟。

Coinbase 场景

可能: User Preference / Configuration / Feature Flag
不是: Transfer / Wallet

四种 Database 对比

Database	最适合	不适合
MySQL	Transaction、Relationship	Full-text Search、Huge Write
MongoDB	Flexible Schema	Strong Relational Data
Cassandra	Massive Writes、Horizontal Scale	Transaction
TSDB	Time Series	General OLTP

Storage Decision Tree

Need Transaction / ACID?            → MySQL
Need Cache / Hot Data / Low Latency? → Redis
Need Event Streaming / Replay?       → Kafka
Need Large Files?                    → Object Storage
Need Full-text Search?               → Elasticsearch
Need Massive Writes?                 → Cassandra
Need Flexible JSON Schema?           → MongoDB
Need Time-series Data?               → TSDB