NoSQL Research
NoSQL Research
The total amount of saved data in the world is rapidly expanding, fueling the need to analyze it all for useful information. The weakness of SQL is that once data is stored, it needs to be rapidly queried and able to scale up easily. NoSQL (Not Only SQL) is the name of the emerging category of databases that do not function the same way as relational databases to store data. Some big changes that they have may include not needing join operations or not having fixed schemas. Popular categories for NoSQL database types include document, key-value stores, graph and column. The common trait is their lack of schema and flexibility to scale horizontally.
In the course MIS 2502, I learned how to use MySQL to construct relational databases. We reviewed how the relational database is comprised of multiple tables to form a schema. Each table has multiple rows of data that can contain items such as customer names, product part numbers, or shipping addresses. NoSQL databases build on SQL databases as some do still recognize querying languages when accessing their data. The fact that there are so many NoSQL options would make it difficult to teach as each company is vying for dominance of the market and there is not one agreed upon query language. The most important point is to still understand how data needs a key as most forms of NoSQL have a way to recall data based on primary key values. Traditional SQL databases may still suit smaller data needs.
Some companies have switched their databases to a NoSQL method. Hulu has chosen to use the column NoSQL product, Apache Cassandra, to store viewer history. The column database uses column families instead of multiple tables as a relational database would. This lets Hulu allow for real-time access to the data from other internal services. Hulu also takes advantage of the multiple datacenter capability of NoSQL by utilizing 16 nodes in two different datacenters. The rapid performance compared to their MySQL server was another factor for Hulu.