NoSQL (not only SQL) or non-relational databases are increasingly being used by financial services businesses and businesses in other sectors where there is a demand for updating data in realtime, and pulling data from different data sources.
The effectiveness of relational databases is often limited by the following factors:
They are designed to run on a single server, and cannot scale horizontally by adding more machines.
Relational databases store data in tables, not objects and relationships between objects. This has the advantage of allowing for batch processing, but requires joins between tables that limit how quickly data requests can be responded to.
SQL databases do not run in memory, and store files on solid state drives.
There are several types of NoSQL databases.
Columnar databases that perform read and write functions on columns rather than rows.
Graph-based databases are designed for data that can be represented in a graph in which different nodes are linked by a finite number of relationships.
Key-value databases that store data as an array of keys and values, with each key only being used once, but having multiple values.
Document store databases that work like key value databases but with the value associated with a key consisting of structured data.
NoSQL databases offer these advantages:
They can guarantee the availability of data at all times, and provide realtime feedback.
They can handle both structured and unstructured data.
They can connect legacy systems with newer systems.
They can easily scale horizontally by connecting to additional servers.
For more information see Data Platforms in Financial Services: The NoSQL Edge, a white paper by MEDICI Global.