DataStax Inc. is hoping to jump-start graph databases out of their niche with a real-time graph engine targeted at cloud applications that manage high-volume changeable data. The company is today introducing an overhauled version of the TitanDB product it picked up with the acquisition of Aurelius LLC 14 months ago that is tightly coupled to the Cassandra NoSQL technology that the company helped develop and which it sells in a commercial distribution as Datastax Enterprise.

DataStax Enterprise Graph is being positioned as a highly scalable and available graph engine that leverages the unique capabilities of Cassandra to support high-volume writes and columnar analytic processing. The distribution uses Apache TinkerPop, a framework for graph databases in both online transaction processing (OLTP) and analytical (OLAP) environments to which DataStax has been the principal contributor. TinkerPop is still in incubation.

“We’re trying to build the first scale-out graph database that works in practice and makes it possible to distribute across multiple data centers,” said Matthias Broecheler, director of engineering at DataStax and a former Aurelius managing partner. “We’re taking on loads that don’t fit in memory, scaling to billions of edges and vertices, all integrated with the search we already have in DataStax Enterprise.” The company is using Apache Spark as the integration layer to bring together DataStax Enterprise’s search capabilities with the graph database’s flexible relationships.

Graph databases have so far been a niche player in the next-generation database world, but the growing popularity of interactive analytics may be setting the stage for growth. The primary strength of graph databases is their ability to accommodate complex and constantly changing relationships with ease. For example, a film database might use one to easily manage relationships between the multitude of people who work across many movies.

“Graph is an excellent method of evaluating, expressing and analyzing previously unrecognized relationships in data,” summarized Gartner Inc. analysts Mark Beyer and Nick Heudecker in a report issued last summer.

The relationship management element makes graph databases popular in applications like recommendation engines, which need to incorporate new data about customer preferences and purchases on the fly. However, graph databases are not considered to be especially good choices for high-volume transactions or those that require a structured query language. DataStax is hoping to change some of those perceptions.

DataStax Enterprise Graph works only with Cassandra, a design decision that Broecheler said provides “major ease-of-use benefits, particularly compared to Titan. We’ve changed the internals to make Cassandra a solid way to traverse the database.  You no longer have to synchronize all the index structures. We can synchronize them so that even if one fails we can recover,” he said. “Our strategy was to pick one database – Cassandra – and tune it to get the maximum power.”

DataStax Enterprise Graph maintains backward compatibility with TitanDB, DataStax said. It takes advantage of Cassandra’s strengths in areas like high availability, high-volume write capacity, linear scalability and rapid response times. The enterprise edition features an adaptive query optimizer, automatic graph data partitioning, a distributed query execution engine and graph-specific index structures. The distribution also ships with the DataStax OpsCenter management suite, DataStax Studio (above) for graph visualization and query development and a set of drivers for popular development languages.

DSE Graph will be sold as an optional add-on to DataStax Enterprise beginning later this quarter. Pricing has not been set.