#🧱 Building Block: Apache Cassandra

#Overview

Apache Cassandra is a distributed, wide-column NoSQL database designed for high write throughput and linear horizontal scalability with no single point of failure.

#Key Concepts

• Wide-Column Store: Data organized by partition key and clustering columns, not traditional rows/columns.
• Distributed Architecture: Peer-to-peer ring topology — every node is equal, no master/slave.
• Tunable Consistency: Per-query consistency levels (ONE, QUORUM, ALL) let you trade consistency for availability.
• Partition Key: Determines which node stores the data; critical for even distribution and avoiding hotspots.
• Clustering Key: Determines sort order within a partition; enables efficient range queries.
• Compaction: Background process that merges SSTables and removes tombstones (deleted data markers).

#When to Use

• High Write Throughput: Event logging, time-series data, IoT sensor data (100K+ writes/sec).
• Time-Series Access Patterns: Transaction history, user activity feeds (partition by entity, sort by timestamp).
• Multi-Region Replication: Built-in cross-datacenter replication for global applications.
• No Single Point of Failure: Financial systems, fraud detection where downtime is unacceptable.

#When NOT to Use

• Complex joins or ad-hoc queries (use PostgreSQL).
• Strong consistency required on every read (use PostgreSQL or DynamoDB with strong reads).
• Small datasets (< 10GB) — overkill operationally.

#Trade-offs

• Pros: Linear scalability, high write throughput, no master node, tunable consistency, excellent multi-DC replication.
• Cons: No joins or aggregations, data modeling driven by query patterns (denormalization required), tombstone management complexity, operational overhead.

#Used In This Repo

• Recommendation Engine (#23) — Interaction event store (50B events/day)
• Fraud Detection (#24) — Transaction record store (2B transactions/day)