Graph databases are gaining attention as enterprises work on their next-generation  artificial intelligence (AI) applications. While still a bit of an outlier, graph-oriented databases continue to find a role in the modern data stack. Using the concepts of nodes, edges and properties rather than tables, graph databases represent relationships and store complex data in a way that relational databases can’t.

With the launch of its Aerospike Graph,  Aerospike is now  firmly in the camp of multimodel database providers. That was an area that once seemed the province of top cloud providers. Original NoSQL advocates like Aerospike are offering such technology to meet emerging needs for data connectedness.

The list of multimodel database makers with graph support includes AWS, DataStax, IBM, Microsoft, MarkLogic (acquired earlier this year by Progress), Oracle and others. Also bringing graphs to the modern data management stack are graph specialists such as Cambridge Semantics, Franz, Neo4j, TigerGraph and others.

Multimodel databases have come to support a Swiss-Army-knife-like array of data tools. That means relational traits have joined the nonrelational methods of NoSQL, that document database traits are co-existing with key-value stores and so on. The multiplicity of database types is one of the clearest signs of an increasingly complex data management stack.

Graph representation provides insightful views of data – linking features and providing a view into data connectedness based on data nodes (also known as vertices) that are connected via edges to other related nodes. That is different from more established relational means comprising data tables of columns and rows.

But developers have come to learn that managing these edge connections can become complex as data volume grows. The work required for graph data processing, sometimes described as “edge traversal,” can take longer than is acceptable in real-time enterprise applications that need responses in milliseconds, according to Aerospike’s Lenley Hensarling. That notion informs Aerospike’s design efforts for the property graph underlying Aerospike Graph, according to the company.

“In fraud detection, identity management and ad tech, graph is being applied more and more. But there has not been a good solution for applying graph in large production workloads,” claimed Hensarling, who is chief product officer at Aerospike.

Hensarling said Aerospike has engineered its new graph database to be capable of millisecond graph queries across trillions of vertices and edges, indicating Aerospike Graph adheres to the widely supported Apache TinkerPop open-source framework for graph databases, with support for other frameworks expected.

“Aerospike Graph capably supports our high volume/low latency demands,” according to Keith Johnson, director of enterprise architecture at customer intelligence provider Acxiom. In a statement accompanying the Aerospike Graph release, Johnson said the database provides exceptional real-time and batch look-up capabilities for multiple edge traversals.

This is an evolutionary step for Aerospike, which has continually expanded on its initial flash-memory-friendly key-value NoSQL store. Early on, it moved to support atomic data consistency to match relational database competitors.

Subsequently, it linked to Spark-based analytics and Kafka streaming. In 2022, JSON document data became a first-class citizen in the Aerospike line, and the company further forged a sourcing deal with Starburst so that Aerospike supports SQL queries – the keystone to relational operations, and the polar opposite to NoSQL. Thus, “NoSQL” is not so great a describer of what Aerospike does now.

The company’s approach to Aerospike Graph targets its core user’s quest for ultrafast graph queries for high-volume web-based applications. Enterprise users are looking for rapid processing to guard against online fraud, manage customer experience or provide audience micro-segmentation for advertising.

Graph databases are ready, in Aerospike’s estimation, for wider use by implementers pursuing what the company calls “the Right Now Economy.” For example, Aerospike lists Experian, PayPal, Snap and Wayfair as customers.

Although a global graph database market estimated at $1.9 Billion in 2021 is expected to grow at 22.5% over the next five years, arguments continue on whether NoSQL graph databases are just “niche curiosities.” But, the mass of new data integrations enterprises now face may pave a path to a larger niche.

The data deluge and graph’s connectedness are keys, according to George Anadiotis, analyst, consultant, writer, and founder of Linked Data Orchestration.

“Initially, graph was just another NoSQL playbook, a newcomer, and people were not familiar with it and what it can do. But what's really powering growth is that people get that  it's not actually the volume of your data that matters so much, it’s the connections,” Anadiotis said while participating in a panel  on graph database futures at Aerospike Summit 2023.

“Graph is the best way that you can leverage connections. That's what's fueling growth,” he said. That growth could be further fueled by an expanding role for graph and NoSQL technology in AI-based machine learning systems, including new ones encompassing emergent generative AI.