So far, you've got a clear idea on schema-free NoSQL data store for the model highly distributed computing application. Let us take a deep dive into the various types of NoSQL data store. Broadly speaking, a NoSQL data store is categorized into four areas, namely:
- Key Value
- Column
- Document
- Graph
A few samples of these four categories of NoSQL data stores are depicted in the following image:
Let us explore the four types of NoSQL data stores:
- Key Value: As per the principles of data management, Key is a primary value used to seek the content of the associated record. As self-defined, Key Value databases pair keys to values. A classic example is employee ID, which is the key element to map the employment details of an associate in a firm. Well-known examples of Key Value NoSQL databases are Microsoft Azure Table and Amazon Dynamo DB.
- Column: In a traditional RDBMS, an entire record is persisted in the underlying disk at the table-row level. It helps the RDBMS table to retrieve a particular row faster. Therefore, RDBMSs are termed as row-based databases. In contrast, the modern column-family NoSQL databases store all the values of a particular column together on-disk. Apparently, it makes data retrieval of a large amount of a specific attribute faster. As a result, this approach helps the end user to aggregate queries and analytics scenarios in which the intention is to execute the data retrieval based on a particular field of the record. This type of NoSQL is the perfect solution for any query-based modern distributed application. In my own experience in the development of financial products, I had an opportunity to swim the depth and breadth of column family stores like Cassandra and Vertical.
- Document: By design, these types of NoSQLs consider documents as a record of its storage. In simple terms, a document is generally defined as a grouping of key-value pairs (as referred in the first category of NoSQL). The major difference lies in the Value section of a Document NoSQL on comparison with a Key Value NoSQL. Actually, the document value can be nested to arbitrary depths, which leads each document to carry its own schema. The most popular Document NoSQL industry database is MongoDB.
- Graph: Graph data structure is the fundamental design here. As we know, Graph is a network model with a unique feature to traverse more than one path using uni- and bi-directional paths/edges between nodes. Connections between nodes are termed as edges in graph theory. The strength of a graph database is in traversing the connections between the nodes. However, they generally require all the data to fit on one machine, limiting their scalability. Graph NoSQL is the best suitable data structure in the modern social computing connectivity model. In fact, LinkedIn connections (with all permutations and combinations) are derived using the industry popular Neo4j graph database.