What are NoSQL databases?

Main types of NoSQL databases

There are several main types of NoSQL databases, each with its own specific features and applications:

• Document Databases: Store data in the form of documents, usually in JSON, BSON or XML format. Each document can have a different structure (flexible schema). They are good for storing complex, semi-structured data. Examples: MongoDB, Couchbase, ArangoDB.
• Key-Value Stores (Key-Value Stores): The simplest model, storing data as pairs of a unique key and corresponding value. Provide very fast reading and writing of data by key. Ideal for caching, storing user sessions or profiles. Examples: Redis, Memcached, Riak KV.
• Column-Family Stores: Store data in columns instead of rows. This allows efficient reading of only selected columns for large data sets. They scale well horizontally. Used frequently in big data analytics and systems requiring high write throughput. Examples: Apache Cassandra, Apache HBase, ScyllaDB.
• Graph Databases (Graph Databases): Designed to store and process graph-structured data consisting of nodes (nodes), edges (edges) and properties (properties). Ideal for modeling complex data relationships, such as social networks, recommendation systems, fraud detection. Examples: Neo4j, JanusGraph, Amazon Neptune.

Key features and advantages of NoSQL databases

Compared to relational databases, NoSQL databases are often characterized by:

• Schema flexibility: The ability to store data with different structures without predefining a rigid table schema. This facilitates rapid application development and adaptation to changing requirements.
• Horizontal scalability: The ability to easily add more servers (nodes) to a cluster to increase capacity and performance (sharding), which is more difficult in traditional relational databases (which more often scale vertically – by adding power to a single server).
• High availability and fault tolerance: A distributed architecture of multiple NoSQL databases provides data replication and automatic failover in case of single node failures.
• Performance for specific applications: Various NoSQL data models are optimized for specific operations (e.g., fast key-by-key access, graph linkage analysis), which can provide higher performance than universal relational databases.

Disadvantages and limitations of NoSQL databases

NoSQL databases are not without their drawbacks:

• Data consistency model: Many NoSQL databases use the so-called eventual consistency model instead of the strong ACID (Atomicity, Consistency, Isolation, Durability) consistency typical of relational databases. This means that after writing data, a read from another node may briefly return an outdated version. This may not be acceptable in some applications (such as financial transactions).
• Less mature tools and standards: The ecosystem of administrative, monitoring and analytical tools for NoSQL databases is often less developed than for mature relational databases. Lack of a single standard query language (like SQL).
• Management complexity: Managing a distributed cluster of NoSQL databases can be more complex than administering a single relational database.
• The need to choose the right model: Choosing the wrong type of NoSQL database for a given problem can lead to performance or complexity issues.

When to use NoSQL?

NoSQL databases are a good choice when dealing with: very large data sets (big data), the need for high scalability and availability, variable or unstructured data, the need for rapid application development or specific performance requirements for specific operations (e.g. caching, graph analysis). It is often used in conjunction with relational databases (Polyglot Persistence approach), using each type of database for what it is best at.