Types of Databases in System Design

Databases are among the most important components usually implemented in a system since they provide for the storage and the management of data. Selecting a database has a pronounced effect on the system’s performance, scalability, consistency, and availability. Choosing the proper type of database for a particular program will achieve an effective operation, thus impacting the efficiency of the user interaction and the speed of the system.

Types of Databases in System Design

Below are the most common types of databases in system design:

1. Relational Databases (RDBMS)

Relational databases in system design are structured storage systems that organize data into tables, each with predefined columns and rows. They ensure data integrity and facilitate efficient querying through Structured Query Language (SQL).

• The relational model supports relationships between tables using primary keys (unique identifiers) and foreign keys (references to primary keys in other tables), enabling complex data retrieval and manipulation.
• This model is widely used in applications requiring structured data storage and ACID (Atomicity, Consistency, Isolation, Durability) properties, crucial for transactional reliability.

Some of today’s well-known types include MySQL, PostgreSQL, Oracle and Microsoft SQL Server.

2. NoSQL Databases

NoSQL databases are designed to handle large volumes of unstructured, semi-structured, or polymorphic data types more efficiently than relational databases. NoSQL stands for "Not Only SQL," emphasizing their ability to store and retrieve diverse data formats without rigid schema requirements. These databases are often used in distributed architectures and Big Data applications where scalability, flexibility, and performance are paramount.

They are divided into several types:

• Document Databases: Store data as documents of JSON or BSON. Examples: MongoDB, CouchDB.
• Key-Value Stores: Record data in a storage system where each chunk of data follows the format key-value. Examples: Redis, DynamoDB.
• Column-Family Stores: Prolific storage of data in the columns instead of the rows which is more appropriate in wide-column store cases. Examples: Cassandra, HBase.
• Graph Databases: Most suitable for information that is in relations and networks, Store data as nodes, edges, and properties. Examples: Neo4j, ArangoDB.

3. NewSQL Databases

NewSQL databases represent a class of databases that aim to combine the benefits of traditional relational databases (ACID transactions, strong consistency) with the scalability and performance advantages of NoSQL databases. These databases emerged as a response to the limitations of traditional relational databases in handling modern Big Data and distributed computing requirements.

• NewSQL databases retain the relational data model but introduce innovations in distributed architecture, concurrency control mechanisms, and scalability techniques.
• They are designed to support massive scalability through shared-nothing architectures, sharding (horizontal partitioning of data), and distributed transaction management.
• Unlike NoSQL databases, NewSQL systems maintain ACID compliance across distributed transactions, ensuring data consistency and reliability.

Some of the systems that can be classified under it include Google Spanner, CockroachDB, and NuoDB.

4. Time-Series Databases

Time series databases are specialized databases optimized for storing and querying time-stamped data points or measurements. They excel at handling large volumes of sequential data generated over time, such as sensor data, financial market data, IoT telemetry, and log data from applications or systems.

Key characteristics of time series databases include:

• Time-stamped Data: Data points are indexed and organized based on timestamps, allowing efficient insertion and retrieval of data in chronological order.
• Optimized Storage: Time series databases typically employ specialized storage structures and compression techniques to efficiently store and manage sequential data, minimizing storage footprint while maintaining fast query performance.
• Time-based Queries: They support time-centric queries, such as retrieving data points within a specific time range, aggregating data over intervals (e.g., hourly, daily), and performing time series analytics (e.g., calculating moving averages, detecting trends).

Examples include: InfluxDB, TimescaleDB.

5. Object-Oriented Databases

Object-oriented databases (OODBs) are databases that store data in the form of objects, akin to object-oriented programming concepts. Unlike relational databases that store data in tables, OODBs directly store complex data structures as objects, along with their attributes and methods.

Key characteristics of object-oriented databases include:

• Objects: Data is represented as objects, which encapsulate both data (attributes or fields) and behaviors (methods or procedures). Objects can have complex structures and relationships, mirroring real-world entities more closely than relational databases.
• Inheritance and Polymorphism: OODBs support inheritance, allowing objects to inherit attributes and behaviors from other objects (classes). Polymorphism enables objects of different classes to be treated uniformly through interfaces or shared behaviors.

Examples include: ObjectDB, db4o.

Comparison of Different Databases

Below are the comparison for different databases:

Scenarios for When to Choose Which Database?

Below are the different databases and some scenarios with example of when to choose this database:

1. Relational Databases (RDBMS)

• Scenario: An e-commerce platform that need scomplicated query, business transaction and possess good consistency.
• Example: Including managing of order through PostgreSQL, managing customers through PostgreSQL and managing stock through PostgreSQL.

2. NoSQL Document Databases

• Scenario: A CMS in which flexible schema requirements are required.
• Example: To store articles, blog posts, and contents to be mined from users, MongoDB is employed.

3. NoSQL Key-Value Stores

• Scenario: An instance of the session storage of a web-application where frequent read and write operations are needed.
• Example: Caching the user sessions and configurations to the Redis.

4. NoSQL Column-Family Stores

• Scenario: This real-time, high-speed analytics system is responsible for processing big data.
• Example: Applying of Cassandra system for storage and processing the results of the users’ interactive sessions.

5. NoSQL Graph Databases

• Scenario: A social network application that seems through which people are connected.
• Example: The handling and storing of connection between the users through Neo4j.

6. NewSQL Databases

• Scenario: An international financial environment with strict adherence to the framework’s conformance and a high degree of adaptability.
• Example: Applying Google spanner to control the financial transactions across Multiple regions.

7. Time-Series Databases

• Scenario: A system that consumes large amounts of time-series data originating from various sensors and is able to query it.
• Example: Storing temperature and humidity values into the database and analyzing it with InfluxDB.

8. Object-Oriented Databases

• Scenario: A cad application which deals with multiple large objects and their links.
• Example: Applying ObjectDB to manage and persist the designs and models of engineering.

Conclusion

The type of database to be used in system design is a very critical decision parameter that determines the performance, scalability, consistency and availability of the end product. There are definite advantages and disadvantages to using each database type, and equally obvious is a list of situations, where each type of the database is most effective. This means that when selecting the database, the tradeoffs involved and the requirements of the application will lead you to the right decision.