A new version of NoSQL database Cassandra has been released. The new release has improved support for dense cluster, simplified application modeling, and improved data cell handling.

Cassandra is a distributed column-oriented NoSQL database that was originally developed at Facebook as a NoSQL system. It was then passed to the Apache Software Foundation, and has an impressive list of companies using it including Adobe, Cisco, Digg, Disney, eBay, IBM, Netflix, Rackspace, Reddit, Spotify, Twitter, and the US Government.

Cassandra supports replication across datacenters, and lets you choose between synchronous or asynchronous replication for each update. The data model combines column indexes log-structured updates, strong support for materialized views, and built-in caching.

The database now has support for clustering across virtual nodes, inter-node communication, atomic batches, and request tracing. The introduction of support for virtual nodes (vnodes) is designed to simplify cluster management, and should also improve performance when adding and rebuilding nodes. Atomic batch support means you can ensure integrity across transactions.

The new version also includes CQL3, version 3 of the Cassandra Query Language that promises to overcome design limitations through more natural representation. The major change to CQL3 is the addition of support for multi-column primary keys, which the team is encouraging the use of to make denormalization simpler.