Why It's Hard
High cardinality introduces millions of unique values per dimension.
High dimensionality adds dozens or hundreds of attributes per event.
Together, they create combinatorial pressure that breaks traditional databases through index explosion, unstable query planning, and rising infrastructure costs. Teams are forced to downsample, pre-aggregate, or discard data just to keep systems running.
Why CrateDB is different
When high cardinality and high dimensionality combine, many systems hit a performance cliff.
CrateDB’s columnar approach stores and scans each dimension independently, so the cardinality of one attribute does not inflate memory usage or index structures for others. New dimensions can be added without destabilizing ingestion or query performance.
This enables:
-
Granular analytics across millions of users, devices, or assets
-
Fast root-cause and anomaly analysis at full data resolution
-
Ad hoc analytics without pre-aggregation or schema redesigns
-
Analytics-ready data for real-time dashboards and AI pipelines
Distributed, columnar architecture
CrateDB uses a shared-nothing, distributed design with columnar storage. Data is automatically distributed and queried in parallel, keeping performance stable as cardinality grows.
Automatic indexing
High-cardinality columns and frequently queried dimensions are indexed automatically at ingestion time. No index planning, no reindexing jobs, and no manual tuning as data evolves.
SQL across all dimensions
Query any combination of dimensions using standard SQL:
-
Filter by millions of unique values
-
Group by high-cardinality attributes
-
Combine structured and semi-structured data
-
Run aggregations, joins, and search in one query
Execution plans adapt automatically as query patterns change.
Schema flexibility for evolving data
Add new dimensions instantly using dynamic columns and JSON support. Evolving data models do not require schema migrations or downtime.
