Cassandra Documentation



Apache Cassandra is an open source, distributed, NoSQL database. It presents a partitioned wide column storage model with eventually consistent semantics.

Apache Cassandra was initially designed at Facebook using a staged event-driven architecture (SEDA) to implement a combination of Amazon’s Dynamo distributed storage and replication techniques and Google’s Bigtable data and storage engine model. Dynamo and Bigtable were both developed to meet emerging requirements for scalable, reliable and highly available storage systems, but each had areas that could be improved.

Cassandra was designed as a best-in-class combination of both systems to meet emerging largescale, both in data footprint and query volume, storage requirements. As applications began to require full global replication and always available low-latency reads and writes, it became imperative to design a new kind of database model as the relational database systems of the time struggled to meet the new requirements of global scale applications.

Systems like Cassandra are designed for these challenges and seek the following design objectives:

  • Full multi-master database replication

  • Global availability at low latency

  • Scaling out on commodity hardware

  • Linear throughput increase with each additional processor

  • Online load balancing and cluster growth

  • Partitioned key-oriented queries

  • Flexible schema


Cassandra provides the Cassandra Query Language (CQL), an SQL-like language, to create and update database schema and access data. CQL allows users to organize data within a cluster of Cassandra nodes using:

  • Keyspace: Defines how a dataset is replicated, per datacenter. Replication is the number of copies saved per cluster. Keyspaces contain tables.

  • Table: Defines the typed schema for a collection of partitions. Tables contain partitions, which contain rows, which contain columns. Cassandra tables can flexibly add new columns to tables with zero downtime.

  • Partition: Defines the mandatory part of the primary key all rows in Cassandra must have to identify the node in a cluster where the row is stored. All performant queries supply the partition key in the query.

  • Row: Contains a collection of columns identified by a unique primary key made up of the partition key and optionally additional clustering keys.

  • Column: A single datum with a type which belongs to a row.

CQL supports numerous advanced features over a partitioned dataset such as:

  • Single partition lightweight transactions with atomic compare and set semantics.

  • User-defined types, functions and aggregates

  • Collection types including sets, maps, and lists.

  • Local secondary indices

  • (Experimental) materialized views

Cassandra explicitly chooses not to implement operations that require cross partition coordination as they are typically slow and hard to provide highly available global semantics. For example Cassandra does not support:

  • Cross partition transactions

  • Distributed joins

  • Foreign keys or referential integrity.


Apache Cassandra configuration settings are configured in the cassandra.yaml file that can be edited by hand or with the aid of configuration management tools. Some settings can be manipulated live using an online interface, but others require a restart of the database to take effect.

Cassandra provides tools for managing a cluster. The nodetool command interacts with Cassandra’s live control interface, allowing runtime manipulation of many settings from cassandra.yaml. The auditlogviewer is used to view the audit logs. The fqltool is used to view, replay and compare full query logs. The auditlogviewer and fqltool are new tools in Apache Cassandra 4.0.

In addition, Cassandra supports out of the box atomic snapshot functionality, which presents a point in time snapshot of Cassandra’s data for easy integration with many backup tools. Cassandra also supports incremental backups where data can be backed up as it is written.

Apache Cassandra 4.0 has added several new features including virtual tables, transient replication (experimental), audit logging, full query logging, and support for Java 11 (full support since 4.0.2 version).