- Relational databases are a better fit for transactional workloads and for performing complex queries. They can scale horizontally using mechanisms such as data sharding but at the cost of requiring additional coordination for executing queries. On the other hand, NoSQL databases are best suited for crunching massive volumes of denormalized data. By design, NoSQL databases can efficiently scale horizontally (even across data centers), with many NoSQL offerings promising a linear increase in query performance as more nodes are added to the cluster. The main caveat of NoSQL databases is that they can only satisfy two facets of the CAP (consistency, availability, and partition tolerance) theorem.
A relational database would be a great fit for systems that perform a large volume of concurrent transactions, such as the ones you would expect to find in a bank. On the other...