What partitioning strategies do NoSQL databases use?

Partitioning splits a dataset into chunks that can be placed on different nodes. It is the mechanism underneath sharding. The strategy decides which rows land together, which directly controls query locality and hotspot risk.

Strategies

• Range: contiguous key ranges, good for ordered scans but prone to hotspots on monotonic keys
• Hash: a hash of the key, good for even distribution but poor for range queries
• List: explicit groups such as region or tenant, good for isolation but can be unbalanced
• Composite: a partition key plus clustering columns, common in wide-column stores for local ordering

CREATE TABLE readings (
  sensor_id text,
  reading_at timestamp,
  value double,
  PRIMARY KEY ((sensor_id), reading_at)
);

SELECT value FROM readings
WHERE sensor_id = 'sensor-42'
  AND reading_at >= '2026-10-01'
  AND reading_at < '2026-10-02';

The partition key should match how the application reads. If most queries fetch one tenant's data, partitioning by tenant keeps those reads local. If queries scan time ranges across all tenants, a time-only partition key may create a hot partition and little parallelism.