In an RDBMS, data is organized by tables that are individually identifiable entities. Within those tables exist rows of a data organized into columns. It is possible to retrieve or update entire tables, individual rows, or a group of columns within a set of rows. In contrast, Riak has a simpler data model in which the Object (explained below) is both the largest and smallest data element. When performing any fetch or update operation in Riak, the entire Riak Object must be retrieved or modified; there are no partial fetches or updates.
Keys in Riak are simply binary values (or strings) used to identify Objects. From the perspective of a client interacting with Riak, each bucket appears to represent a separate keyspace. It is important to understand that Riak treats the bucket-key pair as a single entity when performing fetch and store operations (see: Buckets).
Objects are the only unit of data storage in Riak. Riak Objects are essentially structs identified by bucket and key and composed of the following parts: a bucket, key, vector clock, and a list of metadata-value pairs. Normally, objects have only one metadata-value pair, but when there are more than one, the object is said to have “siblings”. These siblings may occur both within a single node and across multiple nodes, and do occur when either more than one actor updates an object, a network partition occurs, or a stale vector clock is submitted when updating an object (see: Vector Clocks).