Graph processing and graph databases have been with us for a while. However, since their physical implementations are the same for every database in production (Node connected to node, or triplets), there's a perception that data modeling (and data modelers) have no role on projects where graph databases are used. This month we'll talk about where graph databases are a best fit in a modern data architecture and where data models add value.

1. Karen Lopez @datachick #HeartData
Heart of Data Modeling
Graph Databases: Where does the modeling go?

2. Yes, Please do Tweet/Share
today’s event
@datachick #heartdata

3. Karen López
Karen has 20+ years of data and information architecture
experience on large, multi-project programs.
She is a frequent speaker on data modeling, data-driven
methodologies and pattern data models.
She wants you to love your data…
She is loves new tech and gadgets

4. How new tech are you?
...so let’s get to know you….

5. Use Q&A
for
formal
questions
Get them in now!
Use chat
to discuss
with each
other
We have a great
community
Yes!
Slides
Recording
…next week…

6. Plan for Today
Why topic?
Graphy Stuff in Relational World
Another way?
Graph Resources

7. Modern Data
Architectures
will have hybrid
Technologies
WHERE ‘HYBRID’ = ‘SQL and
NoSQL’

8. Clarifying terminology
Graphs Graph Databases Graph Processing
8

9. NoSQL, Not Only SQL
9
Relational Graph Columnar
Key Value
Document
Databases
Column
Family

10. Graph & Hierarchy Concepts
Overview

11. Graph Nodes
Node
Node Node
Node
Node
Node
Node
Node
NodeNode
Node

12. Directed / Undirected
Node
Node
NodeNode
Node
Node
Node
NodeNode
Node

13. What is this structure?

14. Ragged Hierarchies
A hierarchy where
there is variability
in the number of
levels across
branches.
Node
Node
Node
Node
Node
Node
Node Node
Node
Node
Node
Node
14

15. Automobile
Engine
Fuel Line Valve Injector Fan
Fan Blade
Bearing
Bolt
Fanbelt
Entertainment
System
Bolt
Radio
Satellite Radio
Media Player
Backup camera
Automobile
Engine
Injection System
Fuel Line
Valve
Injector
Fan
Fan Blade
Fan Bearing
Fanbelt
Entertainment
System
Bolt
Radio
Satellite Radio
Media Player
Backup camera
Energy Graph
What Happens When…….?
Sometimes we take a group
of “sibling widgets” and
make them a widget just for
them. Think “subassembly”.
Then we have to think of this
new group as a widget.
15

16. How we Model Graph in
Relational
Lots of tricks and tips happening here.

17. Recursive Relationship
Self Join
Recursive Association
Dog Ear / Mouse Ear
Bill of Materials
???

18. Data Model – Hierarchy Recursive

19. Data Model – Hierarchy Recursive
19
What happens
when:
We add a new level?
Take one away?
Promote someone?

20. Relational Performance Tricks
Special data types
Adjacency Lists
Path Enumerations
Closure Tables
Nested Sets
20

21. But remember this?
21

22. Now we have M:N
22
Employee
Reporting Structure

23. Hierarchy vs. Real World
23

24. Recursive
Relationships
24

25. Recursives IRL
Reporting structures
Components
Facilities
Documents
Networks
25

26. SQL Server HierarchyID
26

27. But aren’t relational databases about
relationships?

28. Labeled Property Graph
Nodes have properties
(think key-value pairs)
Nodes have labels
(think meta-data and categories)
Relationships are directed
Relationships have names
Relationships have a start and end
node
Relationships have properties
28
Node
property
property
Node
property
NodeNode
Node
Label
LabelLabel
Label

30. http://neo4j.com/graphgist/8139605

31. Graph Databases
Physical Architecture
31

32. TripleStores
Come from semantic technologies movement
A triple is a subject:predicate:object data structure
Individually triples are semantically poor
En masse they provide rich dataset to harvest knowledge and
infer connections
Use RDF and XML--SPARQL for queries
Ginger dances with Fred
Fred likes ice cream
Karen loves data
32

33. Graph Databases – Neo4j
CREATE (matrix1:Movie { title : 'The Matrix', year : '1999-03-31' })
CREATE (matrix2:Movie { title : 'The Matrix Reloaded', year : '2003-05-07' })
CREATE (matrix3:Movie { title : 'The Matrix Revolutions', year : '2003-10-27' })
CREATE (keanu:Actor { name:'Keanu Reeves' })
CREATE (laurence:Actor { name:'Laurence Fishburne' })
CREATE (carrieanne:Actor { name:'Carrie-Anne Moss' })
CREATE (keanu)-[:ACTS_IN { role : 'Neo' }]->(matrix1)
CREATE (keanu)-[:ACTS_IN { role : 'Neo' }]->(matrix2)
CREATE (keanu)-[:ACTS_IN { role : 'Neo' }]->(matrix3)
CREATE (laurence)-[:ACTS_IN { role : 'Morpheus' }]->(matrix1)
CREATE (laurence)-[:ACTS_IN { role : 'Morpheus' }]->(matrix2)
CREATE (laurence)-[:ACTS_IN { role : 'Morpheus' }]->(matrix3)
CREATE (carrieanne)-[:ACTS_IN { role : 'Trinity' }]->(matrix1)
CREATE (carrieanne)-[:ACTS_IN { role : 'Trinity' }]->(matrix2)
CREATE (carrieanne)-[:ACTS_IN { role : 'Trinity' }]->(matrix3)
http://neo4j.com/docs/stable/cypherdoc-movie-database.html

34. Notice anyting?

35. Querying Neo4j with Cypher
MATCH (you {name:"You"})
MATCH (expert)-
[:WORKED_WITH]-
>(db:Database
{name:"Neo4j"}) MATCH
path = shortestPath( (you)-
[:FRIEND*..5]-(expert) )
RETURN db,expert,path

36. IBM Graph

37. Querying IBM Graph
var url = process.env.graphDBURL + '/gremlin';
var query = "def g = graph.traversal(); g.V().has('code','" + req.body.orig +
"').out('route').has('code', '" + req.body.dest + "')";
var opts = { auth: { user: process.env.username, pass: process.env.password
}, json: { gremlin: query } };
request.post(url, opts, function(error, resp, obj) {
var result = (obj.result && obj.result.data && obj.result.data.length >
0) ? obj.result.data[0] : null;
if (result) { // found a route from orig to dest console.log('route
exists from ' + req.body.orig + ' to ' + req.body.dest); } });

38. Another IBM Graph model

39. Titan

40. Tools and Graph Databases
•No native supportERwin
•No native supportER/Studio
•No native supportPowerDesigner
“the data model
is the database”
“the database is
the data model”
ODBC / JDBC connectively for
querying.

41. So what about modeling for graph?
Marketing vs. Real Project
There is a model
The model isn’t the structure
The model would be used to design
the graph(s)
Same modeling issues:
Naming
Properties
Rules
Consistency
Governance

42. Data Modeling & Graph
No* Logical + Physical Data model
The graph is the data model...and the database
Whiteboard data modeling
Traditional data models still have a role

43. Requirements
Data Model
Database*
More
requirements
/ changes /
tuning /
whims
+ Non Model Stuff
Data Model
Driven
Data Model Driven

45. 10+ Tips for Architects
1. Understand the use cases for graph technologies
2. Evaluate/profile your data requirements for
suitability for graph databases and/or graph
processing
3. ACID support varies across products. You’ll want
to test your use cases.
4. Your query data stories will guide your decisions
5. Test your current development tools for support

46. 10+ Tips for Architects
6. Test your database design/data modeling tools
7. Leverage your existing metadata/models
8. True hierarchies are VERY RARE in the real world.
9. Know the questions you have to ask about all the
exceptions
10.Keep asking where the data integrity happens/is
relevant
46

47. Resources
47

48. http://www.neo4j.org/learn/try 48

49. Fun with Graphs
Scotch Whiskeys
Belgian Beer
Bank Fraud Detection
Access Control Management
…and 100 more….
GraphGist Project - http://gist.neo4j.org/

50. Bluemix
http://www.ibm.com/analytics/us/en/technology/cloud-data-services/products/graph.html

51. Trees and Hierarchies in SQL for Smarties
51

52. Whitepaper
http://whitepapers.dataversity.net/content50141/

53. And it’s FREE!
GraphDatabases.com

54. PostgreSQL
Riak
Hbase
MongoDB
Neo4J
CouchDB
Redis

55. “Every design decision
should include cost,
benefit and risk”
- Karen Lopez

56. Thank you, you were great.
Really, really great.
Karen Lopez @datachick
www.datamodel.com