NoSQL company DataStax has acquired the Titan graph database distributor Aurelius.
DataStax engineering vice-president Martin van Ryswyk said customers had been asking the Apache Cassandra distributor for a graph database capability, and that other acquisitions might be in the offing.
“This is our first acquisition and it shows a maturing of the company,” he said.
A graph database is a collection of nodes and edges. Each node represents an entity, such as a person or a company, and each edge represents a connection or relationship between two nodes. Neo4j is another notable graph database and company.
Aurelius founder and managing partner Matthias Bröcheler said: "By combining DataStax’s engineering excellence in distributed database technology with Aurelius’ graph expertise, we will have the ability to build a tightly integrated graph database that will easily scale both horizontally and vertically as business demands and use cases increase.”
Existing customers for Aurelius’ graph database include Pearson Education, which uses it to support customisation of learning in a social networking product, according to Bröcheler.
More on graph databases
“Pearson uses our technology to, for example, identify students who have problems with certain pieces of material in order to suggest alternatives," he said.
"Individuals learn very differently, and personalisation, driven by recommendation engines, can play a role now, with thousands of students online. It’s about applying techniques from online shopping to education.”
Other existing Aurelius customers are Cisco and Los Alamos National Laboratory.
Van Ryswyk said that graph databases are applicable to more than social networking environments and the particular appeal of TitanDB was it scale-out architecture. The supplier will be building a new product, DataStax Enterprise Graph, using TitanDB and the additional input of the Aurelius team. This is expected to be available later in 2015.
DataStax CEO Billy Bosworth said the firm understands the nature of transactional applications has changed greatly in today's "radically connected and highly distributed" world.
“Accordingly, we purpose built our unified database platform to offer businesses the capability to apply the right database technology to the appropriate use case," he said.
"Graph capabilities on top of Cassandra’s power and scale is an exciting step forward for our customers, and we are delighted to bring the brilliant graph minds at Aurelius to the DataStax team.”
Over the last year we got consistent feedback from our customers and prospects that they wanted graph capabilities combined with Cassandra
Martin van Ryswyk, DataStax
Van Ryswyk added that DataStax's customers go the company because of its scale and availability over a clustered distributed format.
"Over the last year we got consistent feedback from our customers and prospects that they wanted graph capabilities combined with Cassandra. They liked Titan and pushed us in this direction,” he said.
DataStax customers will, with the capability added to Cassandra from Titan, be able to respond to threats quicker and will be more operationally efficient, with “fewer moving parts", according to Van Ryswyk.
"And it will enable scale of joins that you could never do with a relational database technology like Oracle’s," he said. "You can scale more efficiently with a graph model. Our customers are some of the world’s largest retailers and internet companies, and they are on the bleeding edge.”
DataStax customers include Netflix, Adobe, Intuit and eBay.
Bröcheler said the question of how graph databases can be scaled out across a distributed network was at the heart of the two-and-a-half-year-old Titan project.
“How you split a graph so that you can distribute multiple components across different machines and how you replicate that data are the two aspects of the problem," he said.
“Also, from an indexing perspective, you have to cope with skewed distribution, such as people on Twitter with millions of followers and others with a mere handful. We’ve built specialised index structures to cope with that problem, together with what we call vertex cut partitioning.”