A bucket is a subset of data in a sharded cache. Sharding divides the cache across multiple nodes so no single node must hold or handle every key. This differs from an Amazon S3 bucket, which is a container for objects. section 11.6.2 section 7.2
A Redis or Memcached client uses a hashing algorithm to choose the node for a key. A shard is stored on a primary node and can also be synchronized to replica nodes. Sharding can increase cache capacity and distribute reads and writes; replication adds nodes that can handle reads. section 11.6.2
Consider sharding when one cache node can no longer handle the data or requests. For a Redis workload dominated by reads, replication is another option. section 11.6.2
