Sharding is a method of splitting and storing a larger dataset in multiple databases. Sharding is necessary if a dataset is too large to be stored in a single database. By distributing the data among multiple machines, a cluster of database systems can store larger dataset and handle additional requests.