1. Neo4j GraphStore for
LinkedData
May 13, 2013 Arka Pattanayak

2. Overview
2
 Connected (or Linked) Data and the NO-SQL
movement.
 Modeling Linked Data:
 Using the Graph data structure.
 Property Graphs.
 Graph-based Database:
 Neo4j
 Querying a graph-store, two schools of thought:
 Traversal-based
 Pattern-matched

3. Linked Data
3

4. Linked Data According to the W3C
 Large scale integration of, and reasoning on,
data on the Web.
 Standardized format:
 Resource Description Framework (RDF)
 Access to data:
 XML, XHTML, etc.
 Published relationships between data.
 Query endpoints:
 RDF, GRDDL, POWDER, RDFa, the upcoming
R2RML, RIF, SPARQL
4
@see: http://www.w3.org/standards/semanticweb/data

5. Modeling Linked Data: Graphs
 “A Graph —records data in→ Nodes —which
have→ Properties”
 Neo4j Graph: An object that contains vertices
and edges.
 Element: An object that can have any number of
key/value pairs associated with it (i.e. properties)
 Vertex: An object that has incoming and outgoing
edges.
 Edge: An object that has a tail and head vertex.
5
@see: https://github.com/tinkerpop/blueprints/wiki/property-
graph-model

6. Modeling Linked Data: Property Graphs
 A property graph has these elements:
 a set of vertices
 each vertex has a unique identifier.
 each vertex has a set of outgoing edges.
 each vertex has a set of incoming edges.
 each vertex has a collection of properties defined by a map from
key to value.
 a set of edges
 each edge has a unique identifier.
 each edge has an outgoing tail vertex.
 each edge has an incoming head vertex.
 each edge has a label that denotes the type of relationship
between its two vertices.
 each edge has a collection of properties defined by a map from
key to value.
6
@see: https://github.com/tinkerpop/blueprints/wiki/property-
graph-model

7. Modeling Linked Data: Property Graphs in
Neo4j
7
@see: https://github.com/tinkerpop/blueprints/wiki/property-
graph-model

8. Modeling Linked Data: Why not RDBMS?
 Property Graph Model vs. RDBMS Anti-patterns.
8
@see: http://www.scribd.com/doc/2670985/SQL-
Antipatterns
@see: https://github.com/tinkerpop/blueprints/wiki/property-
graph-model

9. 9
@see: http://cdn.memegenerator.net/instances/400x/23186449.jpg

10. NO-SQL (Not-only SQL) Movement
 NO-SQL DEFINITION:
 Next Generation Databases mostly addressing
some of the points: being non-relational,
distributed, open-source and horizontally
scalable.
 Began in early 2009 and is growing rapidly.
 Characterized by:
 schema-free, ✔Neo4j
 easy replication support, ✔Neo4j
 simple API, ✔Neo4j
 eventually consistent / BASE (not ACID), a huge
amount of data ✗Neo4j conforms to ACID! 
 …and more.
10
@see:http://nosql-database.org/

11. Modeling Linked Data: NO-SQL
11

12. Neo4j
“An embedded, disk-based, fully
transactional Java persistence engine
that stores data structured in graphs
rather than in tables.”
– Neo Technologies
12

13. Neo4j: Introduction
 Open-source codebase.
 Baked-in licensing flexibility:
 GPL: “If your app is free, Neo4j is free. If not,
there is a fee”.
 Feb 2010 – v1.0 released.
 Neo Technologies
 CEO: Emil Eifrem (@see:
http://www.youtube.com/watch?v=q9m_5xiGrf4 )
13

14. Neo4j: Understanding the Architecture
14
@see: http://docs.neo4j.org/chunked/stable/tutorials.html

15. Neo4j: An Accessible API
15
Updating operations:
Transaction Wrapper (ACID):
@see: http://docs.neo4j.org/chunked/stable/tutorials.html

16. Neo4j: Querying via Pattern Matching
 “A Traversal —navigates→ a Graph; it —
identifies→ Paths —which order→ Nodes”
 Functional front-end manager application:
 HTTP console (uses REST)
 Cypher (a declarative graph query language)
 Gremlin (an imperative, XPath-oriented, turing-
complete graph programming language)
 Result visualizations
 Query framework plugins for specific use cases:
 e.g. SPARQL
16
@see: http://docs.neo4j.org/chunked/stable/tutorials.html

17. Neo4j: Querying with Cypher (in Java!)
17
@see: http://docs.neo4j.org/chunked/stable/tutorials-cypher-java.html

18. Neo4j: The Good
 Server-side Plugins
 SPARQL plugin for Semantic Pattern Matching of
RDF triples.
 Caching plugins
 Visualization plugins
 Big-Data plugins
 many more and growing…
18

19. Neo4j: The Bad
 RESTful calls to standalone server are slow.
 Cypher is a whole new language.
 SPARQL support is in its infancy.
 Scaling up requires some imaginative re-tooling
of the Property Graph model.
 Indexing is limited to Nodes and Relationships.
19

20. Neo4j: Visualization of Node-space Partition
20

21. :::Live Demonstration & Discussions:::
21