Big data startup Splice Machine is taking its Hadoop realtime relational database to the next level, offering it as a free public beta and as a fully supported commercial product.
The much-anticipated software is designed to help businesses get around Hadoop's batch-analytics limitations, providing a full-featured, transactional SQL database on Hadoop that can run operational applications and realtime analytics.
"We're finally making Hadoop realtime," said Monte Zweben, Splice Machine co-founder and CEO, in an interview. "We're really trying to enter the database space with a bang."
One set of Splice Machine target customers are IT organizations that run traditional relational databases, such as Oracle Database, Microsoft SQL Server, IBM DB2, MySQL and PostgreSQL, but they are running into performance limitations as data volumes explode. Other potential customers are companies that have implemented Hadoop to manage their big data, but discover they can't easily make use of the stored information, Zweben said.
Marketing services company San Antonio-based Harte Hanks is one of the 15 companies that has been working with the beta release of Splice Machine's software. While Harte Hanks has an Oracle-based data warehouse, the company needed database technology that was faster and had more capacity, said Rob Fuller, Harte Hanks managing director of product innovation.
"The challenge there is scaling," he said, adding that the database also needed to work with other applications Harte Hanks uses, including Cognos business analytics, Unica marketing campaign management and Trillium data cleansing.
"We've been putting [Splice Machine] through its paces," Fuller said. In the last three months Harte Hanks has been running the database in "pseudo production" for a retailer client and its 18 terabytes of data. He's seen queries that took 183 seconds completed in 20 and ones that took 32 minutes reduced to 9 minutes. The company now plans on adapting the software more widely.
The Splice Machine Hadoop RDBMS incorporates technology from Apache HBase, the open-source, nonrelational database that underlies Hadoop, and Apache Derby, an open-source relational database written in Java with support for SQL and JDBC. The vendor replaced the storage engine in Derby with HBase and redesigned Derby's planner, optimizer and executor to leverage Hbase's distributed computational engine.
Splice Machine said its database offers a 10x performance improvement over traditional databases and meets ACID (atomicity, consistency, isolation and durability) standards that guarantee database transactions are processed reliably.
Zweben said the channel will be "a very critical part of our strategy" going forward, and a number of global and boutique systems integrators, cloud service companies, business analytic software developers and ISVs are currently evaluating the Splice Machine software.
PUBLISHED MAY 12, 2014