The concept behind Splice Machine sounds like a dare: Take the Hadoop NoSQL data store and use it to create a SQL relational database solution that can scale as easily as Hadoop.
Now, after a beta testing period that began in May, Splice Machine 1.0 is available to integrate with traditional Hadoop jobs and backed by a support program that lets enterprises migrate existing workloads from Oracle, MySQL, and other traditional databases.
At its core, Splice Machine fuses two Apache Software Foundation projects: the venerable Hadoop and the Apache Derby database. Derby is written in Java, and Apache is built to run Java workloads at scale; Splice Machine applies the scale of the former to the technology of the latter. The company claims significant speed improvements with this architecture, based on work done with early adopter partners. More important, the system scales out as freely as Hadoop does.
The point is less to retreat from the advantages that NoSQL systems provide than to speed up (and scale out) ACID-transaction RDBMS behaviors. If Hadoop is becoming a standard-issue data repository for enterprise data, the logic goes, it makes sense to work with it.
Splice Machine 1.0 has been outfitted with features reminiscent of a classic RDBMS solution, such as native backup and restore functions, bulk export, authentication against LDAP (including column-level user privileges), a query-performance analyzer, and SQL-2003-compatible analytics functions.
But one eye-opening feature allows Hadoop work to be done in conjunction with Splice Machine, through support for the HCatalog project. This allows several common Hadoop tools -- MapReduce, Spark, Hive, and Pig -- to interoperate with Splice Machine, and integrates applications written for Hadoop using those technologies. (This falls in line with the company's "best of both worlds" pitch for the product.)
When first introduced, Splice Machine faced a major obstacle to adoption: It wasn't a drop-in replacement for existing RDBMSes like Oracle, IBM DB2, Microsoft SQL Server, or MySQL. To get around that, Splice Machine is offering a support program called Safe Journey, which gives Splice Machine customers a migration path from their existing database solutions. It includes features like converting database stored procedures, which must be rewritten in Java to work in Splice Machine.
Splice Machine's roots in a slew of open source projects open up the possibility that a competing Hadoop or RDBMS vendor will re-create Splice Machine's innovations using the same open source bits. The results could then be relicensed more liberally since Splice Machine isn't offered under an open source license.
Granted, Splice Machine has a free version of the product, but it's offered only to companies "less than five years old and less than $10 million in revenues."
This story, "Splice Machine 1.0 offers speedy, scalable SQL on Hadoop" was originally published by InfoWorld.