BACKGROUND IMAGE: iSTOCK/GETTY IMAGES
Amazon Web Services' re:Invent conference this week in Las Vegas saw a deluge of cloud and database announcements from the cloud computing market leader. Among those on the data side was Amazon Neptune, the company's formal entry into the growing field of graph databases.
While the new AWS graph database may have less immediate impact than Redshift, the influential cloud data warehouse ;rolled out at re:Invent five years ago, it does fill a gap that rivals like IBM, Microsoft and others have included in their cloud data portfolios as they play catch-up with Amazon in the cloud.
AWS CEO Andy Jassy told the re:Invent audience that Amazon Neptune is intended to uncover and map connections in data points in a way that eludes traditional relational database software. With graphs, data is stored in sets of interconnected nodes, unlike relational databases that store data in rows and columns.
Graph databases have found increasing use in online recommendation engines, as well as tasks that include uncovering fraud and managing social media connections. Facebook's Friends and Search graphs is among the most vivid examples of uses of the technology.
Jassy said graph databases, along with NoSQL key-value and document data stores, are part of a trend toward multimodel databases that support a variety of data processing methods, particularly in the cloud.
He said Neptune, which for now is available only in a limited preview, supports graphs based on property and semantic models -- these being the two main schools of graph database construction. The AWS graph database will be offered as a managed cloud service, with automatic data backup to the Amazon Simple Storage Service (S3) over three cloud availability zones.
Andy JassyCEO, AWS
"People have used relational databases for everything," Jassy said. But such single-minded reliance on relational databases is breaking down, he contended.
Neptune isn't Amazon's first foray into graph database technology: AWS already offers the ability to store graphs from the open source Titan graph database and its JanusGraph fork in DynamoDB tables via a back-end storage plug-in. DynamoDB is a NoSQL database developed by AWS for which it claims more than 100,000 users. Neptune's introduction marks a full step into the graph market by AWS, though.
Graph adept and less graph adept
The graph database technology that has emerged in recent years comes primarily from smaller players such as Cambridge Semantics, DataStax, Franz and Neo4j. By and large, these companies have welcomed the AWS graph database into their market, saying that its entry could signify validation of their technology niche.
The established relational leaders -- Oracle, Microsoft and IBM -- have come to include some graph support within their flagship SQL databases, and the latter two have even rolled out stand-alone NoSQL graph databases based on open source technology as secondary offerings.
AWS' target with Amazon Neptune is the relational leaders' flagships, which may struggle when processing ever bigger amounts of graph data, according to Doug Henschen, an analyst at Constellation Research. The addition of the AWS graph database correctly identifies an opportunity for replacing graph analysis use cases currently running on less-graph-adept relational databases, Henschen said.
"Oracle, Microsoft SQL Server and IBM DB2 have all added features for graph analysis," he said. "But SQL and extended SQL functions are not as adept as graph databases and graph query languages at exploring billions of [data] relationships."
To Neptune, and beyond
Neptune was just one of many updates that Amazon added to its fast-moving cloud operation. At re:Invent, Jassy described a serverless version of the Amazon Aurora database, which is now in controlled preview. It can be quickly spun up and down, and customers can pay by the second for database capacity when the database is in use, he said.
AWS is also updating the DynamoDB software by adding global table replication that ensures dependable low latency for data access across many cloud regions. Interest in such capabilities has grown along with the expansion of e-commerce around the globe, according to Jassy.
Global replication for cloud databases was among the traits heralded by Microsoft in its recent debut of the Azure Cosmos DB multimodel database, as well as Oracle in the introductory fanfare for its upcoming Oracle Database 18c release, which initially will be available as a cloud service for data warehousing uses.