Graph databases represent one of the fastest-growing areas in the database market. MarketsandMarkets’ report on graph databases predicts that graph databases will grow from $1.9 billion in 2021 to $5.1 billion in 2026. Organizations are recognizing the power of graphs as a better way to store and manage information across the enterprise. They are being used to deal with a number of data situations such as 360-degree views, recommendations, data mesh approaches, and information integration. Given the popularity of these tools, it is important to understand what they are, how they can best be used, what is happening in the market, and where their strengths and weaknesses are.
For more articles about the future of data management in 2022, download Big Data Quarterly: Data Sourcebook (Winter 2021) Issue
Graph databases (aka triple stores/RDF [Resource Description Framework] data models, or property graph storage) are data management systems that store information about digital entities and how they relate to one another. While relational databases use primary and foreign keys and queries to relate information elements to one another, graph databases show explicit relationships. As a result, graph databases naturally store information in groups of three. There is a subject, predicate, and object. The subject is a specific declaration as to the type of relationship. The predicate is a description of what that relationship is. The object is the item that is being related to.
There are three things to note in the way that graphs store information. First, because the type of relationship is explicitly defined, it is easy to quickly query for common relationships. Second, the information is stored in the way people naturally think and ask questions. Third, the subject and object in this example can be anything. The example I used was a product and its part. It could just as easily be a dataset or a field in the dataset. What this means is that graph databases are very good at defining relationships between disparate data and information. They are very good at pulling together information from different siloed applications and making it more meaningful to business users.
Where Graphs Are Being Used
Graph databases can be used for a wide variety of solutions including data mesh architectures (an approach that creates domain-based connections for distributed
datasets for data consumption scenarios such as machine learning, analytics, or data-intensive applications across the organization).
Graph databases serve as the mesh in a data mesh solution. The graph database is
designed with an ontology that models the way the business works. The graph is then mapped to specific data elements using a feature called virtual graphs. Business users can then query the graph without having to understand how the underlying data is structured. The graph understands where the data is and returns it in a format that makes sense to the user.
In addition, graphs can be used as recommendation engines by mixing data and content and using the relationships in the graph to identify products or information that is closely related. Graphs are also being used as aggregation tools. Organizations looking to get 360-degree views of customers, employees, products, services, and important topics are aggregating this information from multiple sources by pulling it into a graph with the customer, employee, product, service, or topic as a node or subject in the graph.
Finally, graph databases are being integrated into company search engines to expose the aggregated 360-degree views, improve search relevance, integrate data into search results, and provide better support for natural language queries such as those found in chatbots. Organizations that choose to invest in graph technology are using all of these different approaches to get great value from those investments.
Getting Started With Graph Databases
Once an organization has decided it needs a graph database, it must decide which type of graph database is best for its needs. The graph database market is divided into two types of products: RDF graphs and property graphs.
RDF graphs are based on the W3C standard resource description framework and they store all information as triples. Property graphs were pioneered by Neo4j and they are designed to store triples as well as properties about each entity (subject or object) in the graph. While Neo4j has the largest market share of any graph database, many organizations prefer to stick with RDF graphs because the standards make it easier to integrate with them and allow for an easier migration should the organization decide to switch graph database tools.
Organizations that choose to use property graphs tend to select them because they are more familiar constructs for their developers to work with and because products such as Neo4j have a number of out-of-the-box visualization capabilities that are very enticing.