daniels-notes

- Advantages
  - Elastic Scalability - easy add or remove nodes
  - Peer to peer instead of master slave => No single point of failure & Write to any node
  - Fault tolerance to failure of individual nodes
  - Great analytics capabilities (e.g. with Hadoop, Spark, ...) [because of column orientation?]
  - Flexible data model (no strict schemas)
  - Fast INSERT because of caching and commit logs
- Disadvantages
  - CQL is SQL but stripped down
  - Doesn't do data validation like NULL-constraint, uniqueness violations, ...
  - Needs repairs sometimes
  - No transactions
  - Updates and deletes are slow (tombstones)
- Keep in mind:
  - Denormalize data => Super fast reads (no joins) but more data duplication
  - CAP: Cassandra is an available, partition-tolerant system that supports eventual consistency

Terminology

| Cassandra          | RDBMS Equivalent |
| ------------------ | ---------------- |
| Keyspace           | Database         |
| Table/ColumnFamily | Table            |