Apache Cassandra

Apache Cassandra is an open source distributed NoSQL database built for scalability and high availability without sacrificing performance, running on commodity hardware or cloud infrastructure and used by thousands of companies for mission-critical data.

The masterless architecture is the design decision everything else follows from, and it is the reason Cassandra exists as a distinct category of database. A system with a primary node has a component whose failure stops writes, however good the failover is, and failover itself is an event with a duration and a risk. Every node in a Cassandra cluster accepts writes, so there is no such component and no failover to get wrong.

The consequence is stated plainly by the project: a cluster will withstand an entire data centre outage with no data loss, across public cloud, private cloud and on premises. Multi-datacentre replication is described as best in class, giving lower latency to users near each region and survival of a regional outage, and failed nodes can be replaced with no downtime at all.

Linear scalability is the other half of the proposition and the reason large deployments choose it: adding nodes adds throughput proportionally rather than hitting a coordination ceiling. The project takes verification seriously, testing on clusters as large as 1,000 nodes against hundreds of real world use cases, which is the kind of evidence that matters when the failure mode being claimed is rare by design. Cassandra 5.0 is the current release. It is free and open source under the Apache Software Foundation, with a Catalyst programme, published case studies and Cassandra Summit talks.