Interest in graph databases has been trending up in recent years, and analysts have predicted that enterprise use of the technology will continue to grow throughout 2021 and beyond. This assessment is backed by research from MarketsandMarkets, which expects the graph database market will see double-digit growth, expanding from $1 billion in 2019 to $2.9 billion by 2024.
This growth is fueled by the advantages graph databases provide organizations to make more sense of the vast volumes of data they’re gathering. Graph databases help determine relationships in data that are more difficult or even impossible to uncover using other technologies.
This is one of the top graph database advantages that appeals to executives as they turn to data to advance enterprise objectives.
“Any organization, no matter what size or kind, can benefit from graph,” said Mark Beyer, vice president and analyst at Gartner.
What is a graph database?
Relational databases tend to be optimized for speed and structure, which means data is represented as a table without too many missing values and with clear rules on what makes for a valid record or data entry.
A graph database uses graph theory to store, map and search relationships. It is made up of nodes and edges. A node represents a piece of data or entity such as a person, place, thing or category. An edge is a connection or relationship between nodes identifying how they interact.
“Graph databases are very good at dealing with data that is relational,” said Mayank Kejriwal, research lead at the Information Sciences Institute at the University of Southern California.
Mark BeyerVice president and analyst, Gartner
A graph database uses a data ingest engine to put data into a graph configuration. Once the data is in the graph configuration, a user can explore and analyze the connections.
“Graph is about showing me all the connections,” Beyer said.
Consider its use in public transportation, said Sripathi Jagannathan, general manager of data engineering and platforms at UST Global. A transportation network is made of different bus stops — nodes — and routes — edges. Additional stops and routes to service a growing population of travelers are also different nodes.
“In such a use case, the data can be analyzed in an efficient manner [using a graph database],” Jagannathan said. “Even as the number of stops, travelers and routes grow, query performance will remain consistent.”
Why the hype?
The potential for graph database and graph analytics is significant, experts said, because they dramatically expand the ability to find connections in large amounts of data.
“A graph database concentrates on all possible combinations of two or more data objects as opposed to a subset of the preferred joins [combinations] of objects in a data set, so with graph database management it’s all versus preferred,” Beyer said.
Although analysts expect significant growth in sales of graph technologies, use remains low currently. Beyer said Gartner puts adoption at 4% to 6%, adding that graph technologies aren’t as business-user-friendly as other database and analytics tools on the market.
He said data scientists remain the primary users of graph databases for now, but advances are changing dynamics.
“Many NoSQL databases have been optimized for scale, with the number of such databases growing after MapReduce/Hadoop became mainstream, while graph databases allow application developers to represent their data in rich ways,” Kejriwal said.
Graph databases were initially not as fast as leading relational or other NoSQL databases, he said. This has changed in recent years, however, as major providers such as Amazon began supporting graph databases, he said.
“While it is likely that graph databases will become — and may already be, for many applications — scalable enough, most NoSQL databases will not be able to offer the representational use and flexibility that graph databases can,” Kejriwal said.
But as data sets become for most applications less structured and more diverse, graph database will continue to thrive over other NoSQL options.
Here’s a look at the top graph database advantages for enterprises.
Graph technologies are particularly useful in discovering minority positions in not just a data set but the entire body of data, Beyer said. Although minority positions can be meaningful and important in multiple ways, other technologies aren’t as capable of finding them.
Data scientists can use graph databases to accelerate finding relationships and patterns in a data set during discovery, experts said. Data scientists can also use graph databases for real-time analyses, even in massive and complex data sets.
“A graph database gives you ideas from the data, whereas with most other databases you have to have the idea and then ask the data,” Beyer said.
Graph database advantages include helping users find customer service problems, supply chain issues or underserved market segments they didn’t know existed.
“It actually helps find things; it says, ‘I’ve got something,'” Beyer said.
The ability to work with information skew
Graph representation also works with diverse and incomplete data, Kejriwal said.
“For example, some individuals may have very complete profiles and lots of links, while others may have sparse representations,” he said.
Other databases may struggle to deal with this kind of information skew. Graph databases provide a powerful representational mechanism that lets developers represent diverse entities and relationships in an intuitive way, Kejriwal said.
“While NoSQL can address some of these challenges, they are not able to represent and provide querying capabilities over relational degree in the rigorous ways that graph databases can,” he said.
“Hierarchical, relational and document databases suffer from model rigidity, which means the underlying model based on which the data is stored is predefined,” Jagannathan said.
One of the top graph database advantages is the way they represent data. The lack of rigidity allows more agility in how data is stored, and relationships are developed between data points.
“[Nodes] can be stored independent of their edges or relationships,” Jagannathan said. “The relationship can be inserted when one develops. This key feature gives graph databases several other advantages like performance, agility and flexibility.”