NoSQL Databases

NoSQL Features

• Support large volumes of data
• Can be used for data of record or big data analysis
• Non-relational (tables cannot have foreign key constraints)
• Easy to distribute data amongst multiple nodes (cluster-friendly)
• No set schema - even within one table each row may contain data with different attributes (columns)
• Popular database: mongoDB, Cassandra, CouchDB, RavenDB, Neo4j, HBASE, redis, riak, Project Voldermart, DynamoDB, Azure Tables

Data Models

Key-value

• Given a key, give me the value. The value can be anything. Persistent hash map.
• The value content cannot be queried directly but can store metadata and write queries against the metadata.
• riak, redis, Project Voldemar, Dynamo

Document Data Model

• Storage of large number of documents; each document is a data structure, usually JSON
• The contents can be queried
• mongoDB, RavenDB, couchDB

Column-family

• Key mapped to a one or more column families. Each column family has key-value pairs. Multiple families (e.g. Invoice and InvoiceLineItems in one record).
• Conducive to distribution of data amongst nodes (don't have to jump nodes to get one aggregate)
• Entire aggregate in one record.
• Cassandar, HBASE 

Graph 

• NoSQL works well when accessing a pre-defined aggregate. To access (re-arrange) a different aggregate is more difficult in NoSQL than RDBMs
• Special queries for graph structures
• Node and arcs structure. Each to move from relation to relation.
• Neo4j