When and Why Choose a NoSQL Database?
Know how it differs from a SQL database
Introduction
A relational database was the common choice for data storage for a long time. As an architect, your only option was, which relational database to use, and often, this was already predetermined by your company.
After such a long period of the relational database, now may be the time for something new — namely, non-relational databases. And why this could be — we want to dig deeper in this post.
Memory
Before we delve deeper into (non)-relational databases, we should first look at different memories.
Usually, there are two areas of memory in computer architectures. One is fast volatile and the other is larger and slower.
The fast one is called “main memory” and is limited in space. In addition, this has the problem that as soon as the power is gone, the data is lost. For that reason, we write data into our larger and slower storage, called “backing store” (Chapter 1. Why NoSQL?)[2].
The backing store can take different forms, and data can be organized in variant ways. For example, in most user applications (like Word), the file system of the operating system takes care of managing the data. However, a database often presents the backing store in enterprise applications. Databases are more flexible than a file system in storing extensive data and are faster in providing data.
What is a SQL database?
Relational databases support using a specific programming language called Structured Query Language (SQL). This language is used for querying and manipulating data in the environment of relational databases.
It is important to note that SQL is not the only language to query over a relational model, but it is the most common [3].
The data is stored in the form of rows and tables. These tables consist of columns (attributes) and rows (records). At the same time, the entries have logical relationships.
A relational database must have four specific characteristics. These characteristics are usually referred to as "ACID" [Atomicity, Consistency, Isolation, and Durability]. For more information, take a look here.
What is a NoSQL database?
The name No-SQL or non-relational refers to the fact that this form of database stores data in an unstructured way, which differs significantly from relational databases, which stores data in tables and rows.
The underlying technology has been around since the 1960s. It is becoming more and more popular these days due to the growing and changing data [2].
Non-relational databases help developers quickly create database systems and store new information. They follow the "CAP-Theorem" [consistency, availability, and partition tolerance], which states that, in every database, only two of the above characteristics can be guaranteed.
What are the differences between SQL and NoSQL?
There are five main differences:
- Language
- Scalability
- Structure
- Characteristics
- Support and communities
1. Language
SQL has existed for over 40 years and is known by many. It is a safe and versatile language but restricts the user to the defined table schema. This is why it is more elaborate to organize and understand the data before consuming it.
NoSQL-Databases are based on a dynamic schema, allowing us to present alternative structures. Therefore offers us greater flexibility and less planning effort. In addition, we have greater freedom when adding new fields and new attributes. It is also possible to use different syntax in different databases.
Unlike SQL database, NoSQL does not have a unified standard interface, so more complex queries can be challenging to execute. There are many dialects of SQL, but most use standard syntax and similar grammatic.
2. Scalability
Many SQL databases scale well vertically, that is, by adding more or better hardware. Due to the different architecture of NoSQL databases, they scale better horizontally. A large number of NoSQL databases work according to the master-slave architecture.
SQL databases can also scale horizontally. But, this is not supported as well.
3. Structure
One of the most significant differences is probably in the data structure. As already mentioned above, SQL-Database-schema often represents relational, tabular data. They contain tables with columns (attributes) and rows (entries).
NoSQL databases aren't required to hold on to a specific format. There are four types of NoSQL Databases: Document databases, Key-value stores, Column-oriented databases, and Graph databases (for more information, take a look here).
4. Characteristics
Just a short recap because I mentioned these already above. NoSQL databases are based on the "CAP-Theorem." In comparison, the characteristics of SQL Databases are referred to as "ACID."
5. Support and communities
SQL-Databases have large communities and defined standards. There are many guides on the world wide web for beginners. Also, SQL-Languages are proprietary or associated with an extensive individual provider.
While NoSQL communities benefit from open systems, usually, the communities are smaller and more splintered. Compatibility with NoSQL databases varies more than with SQL databases. For that reason, you have to look more closely for dependencies.
When to choose NoSQL over SQL?
As you may have noticed from the above explanation, both database systems are designed for different purposes.
Generally speaking, NoSQL is recommended whenever graphical or hierarchical data is involved. But they are also suitable for very large and rapidly changing data sets and companies fast-growing without a specific data schema.
SQL is better fitted if you store highly structured data with relationships between as well as in systems where consistency is essential. If you are not sure, a SQL database is the better choice.
TL;DR;
NoSQL Databases:
- are open-source,
- non-relational,
- schemaless,
- flexible and suitable fitting for horizontal scaling.
SQL Database:
- are single vendor oriented,
- relational,
- schema-based,
- better for vertical scaling.
