Cardinality measures how many distinct values a dataset or column contains. It matters when choosing a key for horizontal partitioning: a high-cardinality column offers more opportunities to spread rows across partitions, while a key with only a few values can leave partitions unevenly sized. section 5.2.1
User IDs versus order status
A user ID column has high cardinality because each ID is distinct. An order-status column has low cardinality because it has only a few possible values, such as pending, shipped, delivered, and canceled. The latter offers fewer ways to distribute rows by key. section 5.2.1
High cardinality is not a guarantee of balanced load. Requests may concentrate on a few hot partitions, and skewed writes may grow some partitions faster than others. Those cases can require further partitioning, extra resources for busy partitions, or rebalancing. section 5.2.1
