Normalization is the process of defining database tables and relationships in a way that maintains integrity and avoids redundancy. In enterprise applications, we have some reports or views that use heavy SQL Joins to bring data in one table view. These joins are quite heavy operations on the database, and highly affect performance.

Denormalization is a technique in which data can be saved in a single flat table, and the retrieval of data is considerably faster than making SQL Joins or Sub queries. If a business application is complex in nature, we cannot definitely convert its database into a Non-Relational schema. But we can implement the CQR pattern, and modify our READ model to store the values in flat tables, and use some replication techniques to store the data from the write model to the READ model using DTS (Data transformation services), and so on.

On the other hand, if the database is not complex, we can use non-relational (NoSQL) databases, thus reducing query execution time and increasing the responsiveness of the application.

Non-relational databases do not ensure data consistency. They keep that data in key/value formats and do not check the value of the key if it's being replicated within the database.

In Azure, we have two options to use non-relational databases, which are as follows:

• Azure SQL storage
• MongoDB