By Neo4j Staff | May 15, 2015
Excerpt from PwC Technology Forecast – Remapping the database landscape
The power of this level of data aggregation is just now becoming apparent. In the biotech industry, specific skills and knowledge are always at a premium. Companies in public health are starting to use graph-facilitated SaaS to solve business problems. Some of the world’s most knowledge-intensive organizations, including multinational banks, media companies, space agencies, and logistics companies, are also using graph databases, and intelligence agencies have been using them for a decade. Others will follow.
The graph store is one of many innovations creating a sea change in database technology. This issue of the Technology Forecast explores the promise and upheaval caused by these new technologies. This article provides a deeper look at graph technology and how it is similar to other NoSQL1 data stores, which are explored in an earlier article.
What is a graph store?
At Zephyr Health, the pivotal technology is a Neo4j graph store used to find and traverse relationships between entities in data originally ingested using MongoDB, a document database. Document databases are useful for unstructured data, but Zephyr Health had troubles with indexing and latency at web scale, which is why it added the graph database.2
The standard corporate data storehouse, the relational database management system (RDBMS), cannot begin to provide the speedy, flexible search support of a graph. An RDBMS needs absolute consistency among its rows and columns. The difference between a “join” in an RDBMS and in a graph store is like the difference between a precision dovetail joint in woodwork and a freeform Tinkertoy construct. Graphs only need to join or connect at a single point to have useful meaning in searches.
Effectively using graph technology poses its own challenges: the technology is relatively new and only recently capable of web-scale integration tasks. To take advantage of the technology, Zephyr Health needed to resolve some scalability and latency issues associated with web-scale graph environments.
A NoSQL graph store contains malleable maps of entities (named people, places, and things) and how they’re related. Entities become the nodes and relationships become the connections (developers call them edges) in these maps, which can take any shape. If you’re modeling your extended organization, for example, the relationships that appear in a native graph store can add to or become your model.
What’s different in a graph store from a database perspective is the sheer volume of connections, or relationships—how people, places, and things relate to one another through those interactions.
If your data is rich, you’ll see lots of relationships between the entities in native graph form. Older database technologies place less emphasis on relationships, resulting in less context. Graphs offer the chance for richer context through more connections and any-to-any data models rather than the usual tabular or hierarchical models. Graphs can store converted tabular and hierarchical document object information, too, but the graph’s uniqueness and power is in its ability to map any-to-any relationships. Relationship richness of this kind boosts the integration potential and the contextual relevance of the data being represented.
Read the full article.