Only a short few years ago storage options were relatively limited; but recently, both the number and
the types of available storage have exploded. Whereas relational databases continue to support the
highest data integrity and high performance, the table-based metaphor of relational database
management systems (RDBMS’s) has been augmented by other forms of storage including binary large objects (BLOBs), simple key-value tables, document databases, graph-based databases, so-called “Big Data” storage, and others. Application developers have a wide variety of data stores to choose from; increasingly, many cloud-based applications use a combination of several types of storage, a pattern which has been termed polyglot persistence.
For example, a typical e-commerce application will usually have a transactional relational database to
track purchases and sales because an RDBMS of this sort has superb integrity; its transactions are
ACID (atomic, consistent, isolated, and durable). However, these capabilities are not required for every aspect of the application: a simple file will suffice for maintaining logs, a Hadoop-based system can be used for log analysis, a document database can be used for maintaining unstructured data such as a product catalog, and so on. These “NoSQL” data stores can support transactions but typically are not as robust (for this purpose) as a relational database. On the other hand, they are ideal for quickly
deriving insights from very large amounts (petabytes or more) of unstructured data, and the cloud
offers vast quantities of storage space that are typically very expensive for enterprises to provide onpremises.
Source of Information : Microsoft Enterprise Cloud Strategy