2. Outline
• Introduction
• The Lack of relationship for RDBMS and NoSQL
• Graph Databases: Features
• Relations
• Query Language
• Data Modeling with Graphs
• Conclusions
2
3. Introduction
• We live in a connected world
• Everything is connected: Social Network, Biology,
Bioinformatics
• The NoSQL databases analyzed store data using
aggregate
• Here we compare graph databases with relational
databases and aggregate NOSQL in storing graph
data
3
4. Three Facts
1. Relational Databases Lack Relationships
1. NoSQL Databases also Lack Relationships
2. Graph Databases Embrace Relationships
4
5. Relational Databases Lack Relationships
• For decades we tried to accommodate connected,
semi-structured datasets inside relational databases.
• But:
– relational databases are designed to codify tabular
structures
– They struggle when modeling ad hoc exceptional
relationships that are in real world.
5
6. Relational Databases Lack Relationships
• Relationships in relational database only mean joining tables
• But we want to model the semantic of relationships that
connect real world
• As outlier data multiplies:
1. The structure of the dataset becomes more complex and
less uniform
2. The relational data becomes more complex and less
uniform (large join tables, sparsely populated rows, a lot
o null values)
6
7. Example of customer-centric orders
• Complex joins
• Foreign key constraints
• Sparse table with null
values
• Reciprocal queries are
costly “What products did a
customer buy?”
7
8. NoSQL Databases Also Lack
Relationships
• key-value, document, or column-oriented store sets of
disconnected documents/values/columns
• One well-known strategy for adding relationships is to embed
an aggregate’s identifier inside the field belonging to another
aggregate
• But this require joins at the application level
• Some NoSQL have some concept of navigability but it is
expensive for complex joins
8
9. Example of aggregate oriented orders
• Some properties are
references to foreign
aggregates
• This relationship are
not first-class citizens
• Are not intended as
real realtionships
9
10. Example of a Small Social Network
• it’s easy to find a user’s
immediate friends
• friendship isn’t always
reflexive
• We can have brute-force
scan across the whole
dataset looking for friends
entries
10
11. Graph Databases Embrace Relationships
• The previous examples have dealt with implicitly connected
data
• We infer semantic dependencies between entities
• We model the data based on this connections
• Our application have to navigate on this flat and disconnected
data, and deal with slow queries.
• In contrast, in the graph world, connected data is stored as
connected data
11
12. Example Social Network
12
• The node user:Bob is a
Vertex with a property
Bob
• We also see relations
which are Edges:
• Boss_of
• Friend_of
• Married_to
15. Consistency
• Since Graph DBs operate on connected nodes, they
could not scale well distributing nodes across
servers.
• There are solutions supporting distribution:
– Neo4j uses one master and several slaves
– OrientDB uses MVCC for distributed eventual data
structures
– TitanDB partition data by using HBase or Cassandra
15
16. Transactions
• Most of the Graph DB are ACID-compliant
• Before doing an operation we have to start a
transaction.
• Without wrapping operations in a transaction we will
get an Exception.
16
17. Availability
• Neo4j from version 1.8 achieves availability by
providing for replicated slaves.
• Infinity Graph, FlockDB and TitanDB provides for
distribute storage of the nodes.
• Neo4J uses Zookeeper to keep track of the last
transaction Ids persisted on each slave node and the
current master node
17
19. Relations
• Relations in a graph naturally forms paths.
• Querying or traversing the graph involves following a
path.
• A query on the graph is also known as traversing the
graph
• As advantage we can change the traversing
requirements without changing nodes and edges
19
20. Relations
• In graph databases traversal operation are highly
efficient.
• In the book Neo4j in Action, Partner and Vukotic
perform an experiment comparing relational store
and Neo4j
20
21. Relations
• In a depth two (friend-of-friend), both relational db and graph
db perform well enough
• But when we do the depth three it clear that relational db can
no longer deal
21
22. Relations
• Both aggregate store and relational databases perform poorly
because of the index lookups.
• Graphs, on the other hand, use index-free adjacency list to
ensure that traversing connected data is extremely fast.
22
23. Relations another case study
• Let us consider the
purchase history of a user
as connected data.
• If we notice that users who
buy strawberry ice cream
also buy espresso beans, we
can start to recommend
those beans to users who
normally only buy the ice
cream.
23
24. Relations and Recommendations
• The previous was a one dimensional
recommendation
• We can join our graph with graph from other
domains.
• For example, we can ask to fine
– “all the flavors of ice cream liked by people who live near a
user, and enjoy espresso, but dislike Brussels sprouts.”
24
25. Relations and Patterns
• We can use relations to query graph-patterns
• Such pattern-matching queries are:
– extremely difficult to write in SQL
– And are laborious to write against aggregate stores
• In both cases they tend to perform very poorly
• In the other hand, graph databases are optimized for
such kind of queries
25
27. Query Language
• Graph DBs support query
languages such as Gremlin,
Cypher and SPARQL
• Gremlin is a DSL for
traversing graphs;
• It can traverse all the graph
databases implementing the
Blueprints
27
28. 1. Indexing: Nodes and Edges
• Indexes are necessary to find the starting node to
being traversal.
• How Indexes works:
– Can index properties of nodes and edges.
– Adds are done in transactions
• Nodes retrieved can be used to raise queries
28
29. 2. Querying In- Out- Relationships
• Having a node we can query both for Incoming and
Outgoing relationships.
• We can apply directional filters on the queries when
querying for relations
29
30. 3. Querying Breadth- Depth-
• Graph databases are really powerful to query for
incoming and outgoing relationships.
• Moreover, we can make the traverser go top-down
or sideways on the graph by using:
– BREADTH_FIRST or
– DEPTH_FIRST
30
31. 4. Querying Paths
• An other good feature of graph databases is
– finding paths between two nodes.
– Determining if there are multiple paths
– finding the shortest path
• Many Graph DBs use algorithms such as the
Dijkstra’s algorithm for finding shortest paths.
31
32. 5. Querying Paths
• Finally, with Graph DBs it is possible to use Match
operator
• The MATCH is used for matching patterns in
relationships
• The WHERE filters the properties on a node or
relationship
• The RETURN specifies what to get in the result set.
32
35. how do we model the world in
graph terms?
• Formalization of the base model
• Enrich the model
• Testing the model
35
36. Formalization of the base model
• Modeling is an abstracting activity motivated by a particular
need or goal
• We model in order to define structures that can manipulated.
• There are no natural representations of the world the way it
“really is,”
• There are just many purposeful selections, abstractions, and
simplifications that useful for satisfying a particular goal
36
37. Formalization of the base model
• Graph data modeling is different from many other
techniques.
• There is a close affinity between logical and physical
models.
• In relational databases we start from a logical model
to arrive to the physical model.
• With graph databases, this gap shrinks considerably.
37
38. The Graph Model
• A property graph is made up of nodes, relationships,
and properties.
38
40. Nodes
Nodes contain properties
•Think of nodes as documents that store properties in
the form of arbitrary key-value pairs.
•The keys are strings and the values are arbitrary data
types.
40
41. Relationships
Relationships connect and structure nodes.
•A relationship always has a direction, a label, and a
start node and an end node—there are no dangling
relationships.
•Together, a relationship’s direction and label add
semantic clarity to the structuring of nodes.
41
42. Relationships: Attributes
Like nodes, relationships can also have properties.
•The ability to add properties to relationships is
particularly useful for:
– Providing additional metadata for graph algorithms
– Adding additional semantics to relationships (including
quality and weight),
– and for constraining queries at runtime.
42
43. Modeling Steps: Outline
• The initial stage of modeling is similar to the first
stage of many other data modeling techniques, that
is:
– to understand and agree on the entities in the domain
– how they interrelate
– and the rules that govern their state transitions
43
44. Describe the Model in Terms of the
Application’s Needs
• Agile user stories provide a concise means for
expressing an outside-in, user-centered view of the
application needs.
• Here’s an example of a user story for a book review
web application:
– AS A reader who likes a book,
– I WANT to know which books other readers who like the
same book have liked,
– SO THAT I can find other books to read.
44
45. Describe the Model in Terms of the
Application’s Needs
• This story expresses a user need, which motivates
the shape and content of our data model.
• From a data modeling point of view:
– the AS A clause establishes a context comprising two
entities—a reader and a book—plus the LIKES relationship
that connects them.
– The I WANT clause exposes more LIKES relationships, and
more entities: other readers and other books.
45
46. Describe the Model in Terms of the
Application’s Needs
• The entities and relationships in analyzing the user
story quickly translate into a simple data model
46
47. Modeling Rationale
• Use nodes to represent entities
• Use relationships both:
– to express the connections between entities and
– to establish semantic context for each entity
• Use relationship direction to further clarify
relationship semantics
47
48. Describe the Model: Guidelines
• Use node properties
– to represent entity attributes, plus any necessary entity
metadata, such as timestamps, version numbers, etc
• Use relationship properties
– to express the strength, weight, or quality of a
relationship, plus any necessary relationship metadata,
such as timestamps, version numbers, etc.
48
49. Modeling Temporal Relations as
Nodes
• When two or more domain entities interact for a
period of time, a fact emerges
• We represent these facts as separate nodes
• In the following examples we show how we might
model facts and actions using intermediate nodes.
49
55. Iterative and Incremental
• We develop the data model feature by feature, user
story by user story
• This will ensure we identify the relationships our
application will use to query the graph
• With the iterative and incremental delivery of
application features we will be a corrected model
that provides the right abstraction
55
56. Data Modeling: Enrich
• The next steps diverges from the relational data
methodology
• Instead of transforming a domain model’s graph-like
representation into tables, we enrich it.
• That is, for each entity in our domain, “we ensure
that we’ve captured both the properties and the
connections to neighboring entities necessary to
support our application goals”.
56
57. Data Modeling: Enrich
• Remember, the domain model is not totally aligned
to reality.
• it is a purposeful abstraction of those aspects of our
domain relevant to our application goals.
• By enriching our domain graph with additional
properties and relationships, we effectively produce
a graph model aligned to our application’s data
needs
57
58. Data Modeling: Enrich
In graph terms, we are ensuring that:
•each node has the appropriate properties
•every node is in the correct semantic context.
we do this by creating named and directed (and often
attributed) relationships between the nodes to capture
the structural aspects of the domain.
58
59. Data Modeling: Test
• The next step is to test how suitable it is for
answering realistic queries
• Also if Graph DB are great in supporting evolving
structures there are some design decisions to
consider
• By reviewing the domain model and the resulting
graph model at this early stage, we can avoid these
pitfalls.
59
60. Data Modeling: Test
• In practice there are two techniques that we can
apply here
• The first, and simplest, is just to check that the graph
reads well.
• We pick a start node, and then follow relationships
to other nodes, reading each node’s role and each
relationship’s name as we go
• Doing so should create sensible sentences
60
61. Data Modeling: Test
• The second one is to consider queries we’ll run on
the graph.
• To validate that the graph supports the kinds of
queries we expect to run on it, we must describe
those queries.
• Given a described query if we can easily write the
query in Cypher or Gremlin we can be more certain
that the graph meets the needs of our domain.
61
63. Avoid Anti-Patterns
• In the general case, don’t encode entities into relationships.
• It’s also important to realize that graphs are a naturally
additive structure
• It’s quite natural to add facts in terms of domain entities and
how they interrelate adding nodes and relationships
• If we model in accordance with the questions we want to ask
of our data, an accurate representation of the domain will
emerge.
63
64. When to Use
• Connected Data
• Routing, Dispatch, and Location-based Services
• Recommendation Engines
64
65. When Not to Use
• When you need to update all or a subset of entities,
for example in analytics
• In situation when you need to apply operations that
work on the global graph
• When you don’t know the starting point of your
query
65
Notas do Editor
For example, users often want to see their order history, so we’ve added a linked list structure to the graph that allows us to find a user’s most recent order by following an outgoing MOST_RECENT relationship. We can then iterate through the list, going further back in time, by following each PREVIOUS relationship. If we want to move forward in time, we can follow each PREVIOUS relationship in the opposite direction, or add a reciprocal NEXT relationship.