Datastore

A datastore is a repository for storing, managing, and retrieving data used by software applications. It encompasses a broad set of storage systems designed to persist data outside of memory, ranging from simple key-value stores to full-fledged database systems. The term is sometimes used interchangeably with database, but in practice a datastore may refer to any persistent store used by an application, including non-relational and specialized storage engines.

Datastores are characterized by data models and access patterns. Key-value stores map keys to values for fast retrieval; document stores hold self-describing documents, often in JSON-like formats; columnar stores organize data by columns for analytics on large datasets; graph stores manage nodes and relationships. Many datastores are distributed and horizontally scalable. They may be schema-less or have optional schemas, and support indexing and querying features. Support for transactions varies; some provide ACID transactions across records, others offer eventual consistency or tunable consistency models.

In cloud environments, datastores are provided as managed services that handle replication, durability, backups, and failure recovery. They often separate reads and writes across multiple geographic regions to improve latency and resilience. Typical considerations include choosing the right data model for access patterns, understanding consistency guarantees, and planning capacity and cost.

Common examples include Google's Cloud Datastore (rebranded as Firestore in Datastore mode), Amazon DynamoDB (key-value/document), MongoDB (document), Redis (in-memory datastore), and Apache Cassandra (wide-column). Different use cases favor different datastores—session storage, product catalogs, user profiles, event logs, or real-time analytics.