Open Source PostgreSQL 8.3 Better Suited For Web Applications

New features include a full text search tool, improved overall throughput, and ANSI standard SQL/XML support.
The 8.3 release of the PostgreSQL open source database moved out of beta into general availability this week as its development team put the finishing touches to performance enhancements, full text search, and more efficient garbage collection.

Sun Microsystems' Rich Green, executive VP of software, said the 8.3 release "is an impressive new release and we encourage customers around the world to explore it." Sun has provided commercial technical support for PostgreSQL, distributes it with its Java middleware, and employs several full time PostgreSQL contributors, including Josh Berkus, in San Francisco. Sun has maintained its PostgreSQL support after announcing the purchase of MySQL AB, maker of the MySQL open source database for $800 million plus an estimated $200 million for stock options.

The features of the 8.3 release became available for beta testing after Labor Day, but have now been finalized. Berkus said they include the full text search tool, TSearch2, now merged into the core database with new dictionaries and languages. Text searching was previously available as an optional add-on module.

PostgreSQL has improved overall throughput by cutting down on the maintenance overhead in its operations, including transaction throughput. The system formerly paused frequently to clean up rows of data that were no longer being used. PostgreSQL generates versions of rows based on user queries and updates, and retains rows that have become outdated in case they are needed for a back up or transaction recovery. Because of that, it frequently re-indexed tables, "which sucked a fair amount of performance out of the database," said Berkus. Excess row maintenance and re-indexing have been eliminated in the 8.3 release, he said.

That means that the database will perform more predictably, regardless of the variety of tasks, while it was previously more difficult to ensure predictable performance. "The bigger message is performance consistency," not database speed up, Berkus said.

PostgreSQL also can gang up similar queries so that more than one can be addressed during a full scan of a database table, another performance enhancer. That means PostgreSQL, when used for data warehouse or other large database purposes, should be able to return answers to queries on a 10 Tbyte system with the same speed that it used to address a 5 Tbyte database, he said.

The PostgreSQL development team also has added ANSI standard SQL/XML support, including the ability to retrieve data with an XML export command. The command converts the sought-after data to XML and can swap it directly with another database seeking XML data, rather than processing relational data through middleware to convert it into XML. The feature will be useful for Web applications that want to retrieve XML text and documents instead of relational data, Berkus said.

Editor's Choice
Cynthia Harvey, Freelance Journalist, InformationWeek
John Edwards, Technology Journalist & Author
Sara Peters, Editor-in-Chief, InformationWeek / Network Computing