Range-based
Split by key range. Range scans efficient. Hot ranges = hotspots. Rebalancing needed for uneven load. HBase, Bigtable, MongoDB.
Advertisement
Hash-based
Hash(key) mod N or consistent hash. Even distribution. Range scans require scatter-gather. Cassandra, DynamoDB.
Advertisement
Directory-based
Lookup service maps key → shard. Flexible + expensive. Directory becomes SPOF/bottleneck. Manual sharding common.