Non-relational systems.
WHAT THE CATEGORIES ARE
- Key-value: fast lookup by identifier
- Document: structured records without a fixed schema
- Wide-column: rows with varying columns, at very large scale
- Graph: entities and relationships
- Time series: measurements over time
WHAT KEY-VALUE STORES SUIT
Caching Session data Lookups by a known key
WHAT DOCUMENT STORES SUIT
Records whose structure varies Rapidly changing schemas
WHAT THEY COST
Queries across documents being less capable Consistency requiring care
WHAT WIDE-COLUMN STORES SUIT
Very high write volumes at large scale, with known access patterns.
WHAT THEY REQUIRE
Designing the model around the queries, since ad hoc querying is poor.
WHAT TIME SERIES STORES PROVIDE
Efficient storage and querying of timestamped measurements, with downsampling and retention built in.
WHAT THEY SUIT
Monitoring, sensors, and metrics.
WHAT GRAPH STORES SUIT
Traversing relationships: networks, recommendations, fraud rings.
WHAT TO ESTABLISH BEFORE CHOOSING ANY
Whether a relational database would serve.
WHY
It usually would, and it is better understood by everyone.
