College 11: Graphs

Question 1

Benefits of graphs

Answer 1

• Capture naturally heteregeneous data
• No specific schema
• Shows how different parts connect to each other
• Hard to model with relational joins
• Very important for more informative insights

Question 2

Property Graph Model

Answer 2

1. Nodes
   * Represent Objects
   * Can be labeled
2. Relationships
   * Relate nodes
   * Have a label
   * Have direction
3. Properties
   * < Name:value > pairs
   * Describe characteristics
   * Can be on nodes or on edges

Question 3

Different graph sorts

Answer 3

Social networks, media networks, information nets, transportation networks, communication nets

Question 4

Neo4J

Answer 4

Neo4J is Whiteboard friendly, Relationship-focused, and Scalable.
Neo4j is a graph database that prioritizes relationships over indexes, storing data natively as nodes, relationships, and properties across separate store files. Nodes and relationships are managed as fixed-size records with specific bits for reclaiming and flags for characteristics like dense connectivity. Relationships in Neo4j have types and are organized in a doubly linked list, connecting nodes through relationship records. Properties are stored as key-value pairs.

Question 5

Gremlin features

Answer 5

• Designed to work with various graph databases and frameworks, providing a powerful and flexible way to query and manipulate graph data
• Java, Rest, Cypher, Gremlin
• Gremplin: Graph traversal language
• g: variable that represents a graph
• Profiles
• Returns always a se

Question 6

Graph Searches

Answer 6

• Breadth first: explore nearest neighbors first (application: find nearest bank branches that have ATM)
• Depth first: explore as far as possible, then backtrack until you can start exploring again (application: trace root of a problem)

Question 7

Degree centrality

Answer 7

Measures number of relationships a node has.
To estimate popularity of some service or identify fraudsters from legitimate users (centrality of fraudsters tend to be higher in order to inflate prices)

Question 8

Community detection

Answer 8

1. Find groups of users such that relationships betwene memebrs of the group are stronger than relationships across groups
2. Help identify clusters of nodes, isolated groups, or network structure

Question 9

Cypher

Answer 9

A pattern matching query language made for graphs: declarative, expressive, pattern-matching.
Example:
//Find all of the co-actors Tom Hanks have worked with
MATCH (:Person {name:”Tom Hanks”})–>(:Movie)<-[:ACTED_IN]-(coActor:Person)
RETURN coActor.name

Question 10

Neo4J Strengths and Weaknesses

Answer 10

Strengths:
* Open-source.
* Scales well with large datasets (up to 34.4 billion nodes and relationships).
* Typeless and schemaless, though type simulation is available through plugins.
Weaknesses:
* Replicates the entire graph rather than subgraphs.
* Graph size is limited to millions despite high scalability.

Question 11

**

Neo4J internals

Answer 11

Data Storage: Separate files for nodes, relationship labels, and properties.
Memory Management: Uses Least Recently Used (LRU) cache.
Transactions: Managed in-memory with write-ahead logs, also acting as lock managers.

Scalability and Performance

Scale: Supports tens of billions of nodes, properties, and relationships due to its efficient data model.
Latency: Low response times due to traversal-based querying.
Throughput: High read and write throughput due to data locality.

High Availability and Load Balancing

High Availability: Multiple synchronized nodes with a write-master and read-slave architecture. Writes through slaves are not immediately synchronized.
**Load Balancing: **Separates read and write operations and supports cache sharding.