Adding NoSQL Persistence to Storm - Real-time Analytics with Storm and Cassandra - page 117

Columnar database fundamentals
One of the most important aspects of getting started with NoSQL data stores is getting to
understand the fundamentals of columnar databases; or rather, let's use the actual
term—column families.
This is a concept that has a variety of implementations in different NoSQL databases, for
instance:
• Cassandra : This is a key-value-pair-based NoSQL DB
• Mongo DB : This is a document-based NoSQL DB
• Neo4J : This is a graph DB
They differ from conventional RDBMS systems that are row-oriented in terms of the fol-
lowing:
• Performance
• Storage extendibility
• Fault tolerance
• Low or no licensing cost
But having iterated all the differences and benefits of NoSQL DBs, you must clearly under-
stand that the shift to NoSQL is a shift of the entire paradigm of data storage, availability,
and access—they are not a replacement for RDBMS.
In the RDBMS world, we are all used to creating tables, but here in Cassandra, we create
column families where we define the metadata of the columns, but the columns are actually
stored as rows. Each row can have different sets of columns, thus making the whole
column family relatively unstructured and extendible.