Buzz grows around graph databases

Graph databases, which store and query connected information, are starting to emerge as a way of dealing with data delivered in a nonrelational format, such as social networking data.

With a graph database, the focus is on the connections between data. “You’re telling the database in advance that things are connected and how, and representing those relationships physically,” as opposed to storing them in tables and relating them through indexes, said Philip Rathle, senior director of products at graph database vendor Neo Technology, at the recent NoSQL Now conference in San Jose, Calif.

But there is still plenty of room for old-fashioned relational databases. A graph database is for special uses. “I definitely wouldn’t recommend a graph database if you have very tabular and well-structured data. Use a relational database for it,” said Emil Eifrem, Neo’s CEO. “But if you have data that is messy, that is complex, that is connected, then a graph database is vastly superior.” Eifrem cited applications such as a social network or fraud detection, where data is connected, as systems that could benefit from a graph database.

Usage of graph databases is increasing, but the technology still represents a “minority sport,” according to analyst Philip Howard of Bloor Research. “Graph databases are critical when the degree of separation [ie, I know x who knows y who is related to z who used to live in the same house as w etc.] between entities becomes too great to handle using conventional technology. Oracle or DB2, for example, can reasonably handle up to three degrees of separation but not the six or seven degrees you need in, say, telco fraud,” Howard said.

Howard noted the limitations of graph databases. “The major limitation is that while these are technically NoSQL databases, in practice they cannot be implemented across a low-cost cluster (at least not a present) but have to run on a single machine, the reason being that performance degrades rapidly across a network,” he said.  “Another potential drawback is that either you have to write your own queries using Java or whatever — which means employing expensive programmers — or you use SparcQL or one of the other query languages that have been developed to support graph databases, but this means learning a new skill.” There are visualization tools available for graph databases but they are relatively immature, Howard added.

Other players in the graph database space include vendors such as Franz and Objectivity. At Franz, Craig Norvell, vice president of sales and marketing, also points out the connectivity benefits and adds that graph databases can be leveraged in the semantic Web. “From our perspective, [a graph database presents] an opportunity to find connections within your data.” 

Neo, meanwhile, cites customers including Adobe and Moviepilot, with Adobe leveraging Neo’s Neo4j database to add social capabilities to its Creative Cloud platform for accessing Adobe Creative Suite applications. Connected data is stored across three continents while maintaining high query performance, according to a Neo representative.

Adobe’s endorsement could be just a start for graph databases. Look for the buzz around them to only grow, given the interest in analyzing data and expanding beyond what is possible with relational databases.

This story, “Buzz grows around graph databases,” was originally published at InfoWorld.com. Get the first word on what the important tech news really means with the InfoWorld Tech Watch blog. For the latest developments in business technology news, follow InfoWorld.com on Twitter.