To generate interest from developers, Boston-based Akiban Technologies has released as open source its flagship database software, called Akiban Server. The company also released a connector for replicating a MySQL database within Akiban, and forged a partnership with platform hosting provider Engine Yard.
"The data that we capture grows more and more complicated. We designed a database that can handle increasing levels of depth and complexity," said Ori Herrnstadt, who is a cofounder and Chief Technology Officer for Akiban.
The company announced the release at the O'Reilly Open Source Conference (OSCON), being held this week in Portland, Oregon.
Through a novel way of grouping data, the Akiban database can complete operations 10 or more times faster than MySQL, the company has claimed. It also provides developers multiple ways for their applications to work with the database.
The company has been working on the technology for the past 18 months, consulting with 120 organizations to understand their database needs, and has offered a commercial version of the database since January. The company is pitching its software first to online businesses that conduct transactions and maintain customer profiles.
"The key advantage of [Akiban's] approach is that it reduces the need for cross-table joins," noted 451 Research analyst Matt Aslett, in a research report. Typically database administrators design their databases through a process called normalization, in which often-duplicated data items, such as zip codes, are consolidated on separated tables to save space. However, as additional tables are added, the time it takes to assemble the answers to queries -- through joins -- can slow.
To minimize this problem, Akiban inserted a new layer into the standard database architecture, something called table groups. A table group is a hierarchical collection of tables. It sits in between tables and the overall schema that defines the database.
With Akiban, each database entry is stored as its own record, within a table group. For an online retailer, for instance, all of a customer's information would be located in a single record, which would include multiple rows containing contact information and orders placed. "Each row remains completely independent," Herrnstadt said. Object oriented programmers can think of groups as objects, though Akiban is not, strictly speaking, an object database, Herrnstadt said. In database speak, the data is stored in the third normal form (3NF), according to database analyst Curt Monash, of Monash Research.
Despite this variance in the architecture, the database offers full support for all SQL commands and possesses the full set of ACID (atomicity, consistency, isolation, durability) properties that ensure database reliability.
At present, the database can not scale out beyond a single server, except in cases of replication. The company will plans to work on a distributed version of the database, however.
To further entice developers over to Akiban, the company offers a tool that mimics MySQL, designed to appeal to current MySql users. MySQL Replication Adapter is a MySQL storage engine for MySQL that replicates MySQL database to an Akiban server. The company hopes organizations will use the engine to set up shadow copies of MySQL databases, and, over time, replace the MySQL databases with Akiban-based databases.
The company is also setting up a partnership with development platform cloud provider Engine Yard, which will offer a hosted version of the database.
The enterprise edition of Akiban Server has been available since the beginning of this year. It includes additional functionality not found in the community edition. The enterprise version costs typically costs between US$20,000 to $40,000 depending on the features used. The open source community edition, released Wednesday, is under the GPL (General Public License), version 3.