Key-Value Stores and NoSQL
Stub. Why key-value/NoSQL stores trade the relational model's structure and strong consistency for horizontal scalability, worked through Cassandra's architecture (ring-based partitioning, tunable consistency) alongside HBase and MongoDB as points of comparison.
NoSQL stores are the applied payoff of the replication/consistency and peer-to-peer material: Cassandra’s data model and consistent-hashing ring architecture look like a structured p2p overlay (Chord’s ring, specifically) repurposed as a database, with per-operation tunable consistency levels that let an application pick its own point on the consistency/availability tradeoff rather than have one imposed. HBase (a Bigtable-style column-family store layered on a distributed filesystem) and MongoDB (a document store) are covered as points of comparison across the NoSQL design space — partitioning strategy, data model, and consistency guarantees.
Outline (TODO — flesh out each)
- Why key-value/NoSQL stores — the relational model’s scalability limits
- Cassandra — data model, ring-based partitioning, tunable consistency
- HBase — column-family model, layering on a distributed filesystem
- MongoDB — document model