VoltDB 3.0 enhances performance, SQL and connection

The developers behind VoltDB, the in-memory relational database, have released the third major version of the database. VoltDB's in-memory architecture allows it to have low latency and high throughput and, in version 3.0, the developers have re-engineered the transaction coordination architecture to ensure that communication between nodes in a cluster is minimised during transaction processing without compromising the ACID nature of the system. The result, say the developers, is sub 2ms latencies in their test system up to 240,000 transactions per second; VoltDB 2.8.4.1's latency shot up to over 15ms at around the 200,000 transactions per second on the same test.

The reduction in communication between cluster nodes also has the added advantage of making the database more cloud friendly and able to withstand poor network latencies or underperforming nodes. Managing the database cluster should also be easier with a new, scriptable, command-line voltadmin interface to assist with shutdown, saving and restoring clusters.

Performance isn't the only enhanced element in VoltDB 3.0 though; the SQL coverage of the database has been expanded to include SQL UNION, LIKE and NOT LIKE, and a full set of column string and number functions which can be used to define indexes. The 3.0 release also adds JSON functionality with a new field() function for columns, which should allow JSON-structured data to be queried within a column.

Getting data out of VoltDB should also be easier with an improved export architecture which can run twenty times faster than previous releases. A new JDBC export connector allows exporting to PostgreSQL, Oracle and MySQL, among others. New import sources include Apache log file data and comma-separated value input files and there is also a new high-performance PHP driver, a first release of a Node.js driver, and a Google Go driver.

VoltDB is available as a GPLv3-licensed community edition or as a commercial, and enhanced, enterprise edition. The community edition's source code is available on GitHub. Binary versions of the enterprise and community edition can be downloaded, with the enterprise edition on a 30-day trial.

(djwm)