Everything you need to know about Splice Machine’s SQL-on-Hadoop database
Originally featured in the Computer business review. Visit the original article here.
by Amy-jo Crowley| 25 July 2014
The benefits, the challenge and competition.
Startup Splice Machine recently rolled out its relational database management system (RDBMS) into public beta.
Introduced in May 2014, the Splice Machine system combines the Java-based Apache Derby relational database with technology from the Apache HBase NoSQL database,
CBR tells you five things you need to know about it.
1. The aim
Splice Machine, founded in 2012, hopes its RDBMS system will attract customers away from traditional vendors, such as IBM, Oracle and Microsoft it claims are experiencing cost and scaling issues.
Anyone with standard IT skills of understanding a SQL database and starting up a Hadoop environment can deploy it with specialised skills and training, according to Splice Machine's CEO Monte Zweben.
"You don't have to be a Java programmer, you don't have to know MapReduce. Any Oracle database person or SQL database person would be able to use Splice Machine," he explained in an interview with CBR.
2. The Benefits
Zweben pointed to Harte Hanks, a marketing services provider for top brands worldwide, which has seen a greater than 10x price and performance improvement with the Splice Machine RDBMS compared to their existing Oracle RAC databases.
"The problem that they experienced was that as they grew with the number of retailers utilising this management service, the whole architecture was being powered by Oracle RAC," he explained.
"And the Oracle system grinded to a halt and they experienced near performance problems."
He added: "We'll see how well we execute but over time I think this is going to be an explosive opportunity. I think the affordable scale-out solution is critical. I think people are going to buy this software in a different way than they used to buy software...there's very little barrier to entry for people to really get a feel for it."
Splice Machine also complies with ACID (Atomicity, consistency, isolation and durability) properties and claims to be the only SQL-on-Hadoop option that can support concurrent reads.
"So we're a different kind of database, a database that would be used for operational and transactional applications," said Zweben.
"One example might be Web personalisation where you have many concurrent users on a website and your needing to update the data that's coming in real-time but at the same time being able to show the right content, motion or offer or ad to that customer at the right time."
Matt Aslett, research director for Data Platforms and Analytics at 451 Research, told CBR that while the company does have the potential to challenge Oracle, IBM and Microsoft, he expects the transactional database market to be dominated by traditional approaches at least for the foreseeable future.
"However, we have already seen Hadoop grab a small but growing slice of the analytics market, and Splice Machine's approach has been validated by the recent emergence of Hewlett-Packard's Trafodion project, which is also designed to support transactional and operational workloads on Hadoop," he explained.
"In addition Splice Machine partner MapR and Hadoop creator Doug Cutting have also begun talking about the potential they see for Hadoop to support transactional workloads. It is still very early stages but we do see that the concept has real potential."
Aslett added that there are also a small number of "forward-thinking" enterprises that are looking to adopt distributed scale-out architecture for their next-generation database platforms.
The main challenge for Splice Machine is making sure that the performance and quality of their software meets "the critical needs of our initial customers".
Zweben said: "I think in terms of the broader challenges it is really is a matter of execution - the need for scale is apparent and everyone is talking about it. This is about bringing the high quality that's easy for the customer to adopt with very little invasiveness that doesn't require the customer to change their skillsets or replace their existing applications."
5. Business Model
Splice Machine is available on a freemium basis for test developments and experimentation with no support other than access to forums and upgrades. To try it out, companies must be less than five years old and have less than $10m in revenues.
For the "Enterprise Edition", Splice Machine has a list price annual licence fee of $5,000 per node per year. It also guarantees to respond problems in as little as one hour from when they are reported.