Stardog

Last Updated: 5th July 2023
Analyst Coverage: Philip Howard and Daniel Howard

Stardog is developed by Stardog Union. It was founded in 2010 based on research conducted at the University of Maryland’s AI lab. The first version of Stardog was made generally available in 2012. The company is backed by venture capital. It is based in Arlington, Virginia. While it does not have other offices, it does have an EMEA representative along with European customers such as Siemens and the Swiss Government. In general, the company focuses on the Global 2000 and it has significant numbers of users in the finance, healthcare, manufacturing, government and technology sectors. Amongst others, these include LGT, NASA, Cisco, Raytheon, Bayer, and Bosch.

The company has multiple technology partners including Cambridge Intelligence (Keylines), Tableau and Tom Sawyer for visualisation purposes, and Solidatus which leverages Stardog for its data lineage product offering.

Company Info

Headquarters: 2101 Wilson Boulevard, Suite 800, Arlington, VA 22201, USA
Telephone: +1 202 408 8770

Stardog

Last Updated: 5th July 2023

What is it?

Stardog presents itself as an enterprise knowledge graph platform, helping you build semantically-rich knowledge graphs to view and explore complex information across multiple, heterogeneous data sources. One of its primary aims is to provide frictionless knowledge engineering that enables mainstream data users, as well as graph aficionados, to benefit from graph technology, while at the same time accelerating knowledge graph time-to-value. This is achieved in part through a variety of low/no-code experiences that provide self-service, collaborative knowledge graph capabilities to all of its users. For example, semantic search is enabled via a visual query builder. Moreover, Stardog is supported by a robust network of technology partners like Accenture, Microsoft and Databricks and corresponding integrations that can serve to augment the platform with analytics capabilities, data storage, data governance, and more. This also allows it to easily slot into existing ecosystems or as part of a data fabric or data mesh architecture.

Under the hood, Stardog is an RDF graph database with strong support for SPARQL and OWL that can be extended to provide labelled property graph capabilities (via RDF* and SPARQL*) with support for Apache Tinkerpop and Gremlin. It embeds the Lucene search engine, its storage layer is founded on RocksDB (a key-value store written in C++), it is ACID compliant, and it supports immediate consistency. Deployment is based on Docker containers and Kubernetes, and it is available as both a managed AWS offering (Stardog Cloud, which includes Free, Essential and Enterprise price plans) as well as on-prem, complete with SSO support, via Stardog Launchpad.

Customer Quotes

“Deploying on Stardog fosters the idea of reusability among the mapping and vocabularies within the database products. Stardog makes large-scale graph data management possible for Springer Nature — and propels our data-modelling and data-integration efforts to a completely new level.”
Marcel Karnstedt-Hulpus
Director Data & Knowledge Technologies DRG

What does it do?

Figure 1 – Stardog user experiences

Stardog features three distinct user experiences, as shown in Figure 1: Stardog Explorer, a query layer designed for visually exploring your knowledge graphs; Stardog Designer, a no-code, visual environment for building and maintaining your knowledge graphs; and Stardog Studio, a code-based IDE (Integrated Development Environment) for writing more complex queries and administering your knowledge graphs. Designer is particularly notable for the degree of automation it provides for building out the semantic models (ontologies) and mappings that underpin your knowledge graphs. Various domain and/or industry-specific templates (called Knowledge Kits) are also provided to facilitate model-building. Likewise, Studio offers features like contextually aware auto-completion that facilitate increased developer productivity, though their effect is somewhat more subtle owing to its code-driven nature.

Figure 2 – Stardog architecture

Stardog’s architecture is illustrated in Figure 2. It has several notable features, including BITES NLP (Natural Language Processing), an extensible document storage system that provides configurable storage and processing for unifying unstructured data with Stardog graphs, and virtual graphs, which allow you to declaratively map structured data into Stardog and query it from there, unifying structured data into the platform as well. In addition, the platform provides out-of-the-box (virtual) connectors for a range of popular data sources, and an SDK is available for users to develop their own connections.

There are some recently added features worth highlighting. These include geo-replication for high availability clusters; industry Knowledge Kits, which offer predefined ontologies, data models, and other assets tailored to specific verticals or use cases; enterprise metadata knowledge graphs, knowledge graphs that span multiple data catalogues and data sources via pre-built integrations with Collibra, Unity, Microsoft Purview and more; and an entity resolution service (in beta as of writing), which enables you to identify and link data representing the same entities across different data silos via unsupervised machine learning algorithms. Notably, Stardog is also developing Voicebox, a natural language interaction interface designed to produce either SPARQL code or a visual, semantic model corresponding to your natural language queries.

Why should you care?

Enterprise knowledge graphs are currently proving very popular, and for good reason. When you have a large amount of data, and when that data is complex, highly interrelated, and spread out over multiple sources, it can be very difficult to grasp its structure and meaning. In other words, to understand it. Knowledge graphs are an excellent way of addressing this issue, and thus leveraging your data as effectively as possible.

Stardog offers some significant advantages when it comes to knowledge graphs. Its extensive reasoning and inferencing, its support for unifying both structured and unstructured data through a combination of virtualisation and materialisation techniques, its strides towards ease of use and a newcomer-friendly user experience, its capacity for integration, its automated model building, and its (upcoming) entity resolution service are all worth noting. In addition, Stardog is built to operate at enterprise scale, and offers high performance on distributed, real-world data. What is more, the fact that it supports Gremlin as well as SPARQL is helpful, as is the fact that it can operate as a labelled property graph, because while information architects tend to like RDF graphs built on open standards, developers often prefer property graphs, and supporting both gives it the best of both worlds.

The Bottom Line

Stardog is a full-fledged knowledge graph platform supported by an equally full-fledged graph database. This means that it is particularly suited to helping you understand large, complex, highly heterogeneous data environments and capitalise fully on the data therein. In short, if you want to build a knowledge graph, Stardog is an excellent means of doing so.