Key-Value Store
A key-value store is the simplest form of NoSQL database, mapping unique keys directly to values with no required schema, optimized for extremely fast lookups, writes, and deletes by key.
Definition
A key-value store is the simplest form of NoSQL database, mapping unique keys directly to values with no required schema, optimized for extremely fast lookups, writes, and deletes by key.
Overview
A key-value store's data model is essentially a giant distributed hash map: given a key, the database returns its associated value with minimal overhead, and there is no concept of columns, joins, or query languages beyond basic get/set/delete operations (and sometimes simple range scans). This simplicity is exactly what makes key-value stores extremely fast and easy to scale horizontally — with no relationships to maintain, data can be partitioned across many nodes by key with minimal coordination. Values themselves are often opaque to the database — they might be simple strings and numbers, or serialized blobs like JSON objects, images, or session data — with the application responsible for interpreting the value's structure. Some key-value stores like Redis extend the basic model with rich in-memory data structures (lists, sets, sorted sets, hashes) and pub/sub messaging, blurring the line with a general-purpose in-memory database. Because of their speed and simplicity, key-value stores are the default choice for caching layers sitting in front of slower databases, for session storage in web applications, and for any workload that is dominated by simple lookups rather than complex queries or joins. They trade away the ability to query by anything other than the key — if you need to search by value or filter on secondary attributes, a document database or relational database is usually a better fit. Popular key-value stores include Redis, Memcached, Amazon DynamoDB (which layers additional indexing on top of a key-value core), and etcd (used for configuration and coordination in distributed systems like Kubernetes).
Key Features
- Maps unique keys to values with minimal overhead, like a distributed hash map
- Supports simple, extremely fast get/set/delete operations by key
- Scales horizontally with ease since there are no cross-key relationships to maintain
- Values are often opaque to the database, interpreted entirely by the application
- Some implementations (e.g. Redis) add rich data structures and pub/sub messaging
- Cannot efficiently query by value or secondary attributes without extra indexing