Organizations are in an era where the variety and volume of data they have access to are increasing everyday. To utilize the available data, there is becoming an increasing number of ways to store the data. These come in the form of NoSQL databases, often interpreted as “not only SQL”, meaning that the databases can sometimes act as relational databases but are not constrained by relationships between stored tables. As a data scientist, it is becoming more likely that you will have to interact with NoSQL databases, so this blog will introduce you to one, Apache Cassandra, and explore why more organizations are using it to store their data.

NoSQL databases are generally open source with more scale-out than scale-up capabilities. They are built on clusters made of commodity-class hardware and utilize replication as the primary means of failover. These are worthwhile features, but they also necessitate a specific look at when it comes time for selection. These are the features that are generally found in NoSQL databases and form minor evaluation criteria for selection.

NoSQL databases pose a particular challenge because companies are increasingly using them to run large-scale Web and cloud applications. NoSQL data stores are well-suited to operational queries, but aren’t optimized for analytics queries. For example, key-value stores are extremely fast when looking up a record for a given key-such as a user profile. But they can’t easily query a subset of profiles or aggregate millions of records. Other NoSQL databases allow customers to store data using an arbitrary set of columns, creating sparse records and a non-symmetric format that requires users to understand the structure of the data before they query it-something called schema on read-which is the opposite of how SQL-based databases and queries work.