Introduction to NoSQL

(Not Only SQL)

By Antonio Castellón :: Multi-Disciplinary Engineer – Computer Science

May, 2015 - for Philip Morris International R&D

Problem : Data Complex

Problem : Data Complex to Model

Problem : Dynamic Data ( Uncertainty )

End User requirements and data itself sometimes generate different types of uncertainty

The NoSQL Jungle

Data – NoSQL – Different implementations

CURRENTLY +150

Data - NoSQL – Comparing data structure

Image from: http://highlyscalable.wordpress.com/2012/03/01/nosql-data-modeling-techniques/

Data - NoSQL – Compare

98% of the business requirements

There is still billions of nodes and relationships

Data - NoSQL – Keys to fit

Key-value store

Column Store

Document Store

Graph Database

Performance High High High Variable

Scalability High High Variable (High) Variable

Flexibility High Moderate High High

Complexity None Low Low High

Functionality Variable(None) Minimal Variable (Low) Graph Theory

Data – Our selection

Graph Databases

Data – Graph Databases – Why?

Flexible data structureDoesn’t matter if the relations will change in the future.

Closer match to business logic

Data – Graph Databases – Why?

Natural query system You tell what you want, not how to get it.

with recursive cluster (party, path, depth)  as ( select cast(@userId as character varying), cast(@userId as character varying), 1  union  (  select (case  when this.party = amc.userA then amc.userB  when this.party = amc.userB then amc.userA  end), (this.path || '.' || (case  when this.party = amc.userA then amc.userB  when this.party = amc.userB then amc.userA  end)), this.depth + 1  from cluster this, chat amc  where ((this.party = amc.userA and position(amc.userB in this.path) = 0)  or (this.party = amc.userB and position(amc.userA in this.path) = 0)) AND this.depth < @depth + 1 ) )  select party, path  from cluster  where not exists (  select *  from cluster c2 where cluster.party = c2.party  and (  char_length(cluster.path) > char_length(c2.path) or (char_length(cluster.path) = char_length(c2.path)) and (cluster.path > c2.path)  )  )  order by party, path; 

SQL = several hours to be executed

VS

START b = node:User(UserId=‘Manolo') MATCH (b) --(friend)--(friendoffriend) RETURN count(friendoffriend)

Cypher Language = 635ms

Data - Graph Databases – Why?

Fits very well with complex data

Data - Graph Databases – Why?

Fits very well with Bio-Informatics

0.9 Billion relationsips

Data – Graph Databases – Why?

Fast Prototyping and developmentWe don’t need to lose too much time to define the schema (fine-grained).

Data - Graph Databases – What is it?

Properties

Labels

Relationships

Data - Graph Databases - Implemented by …

Data - Graph Databases – Top 3

Name API Query Methods Consistency Staff (people) / Community

OrientDB Java Traverser API, Blueprints, Rexster

Own SQL-like Query Language, Gremlin

ACID, MVCC 3 / Low

Neo4j Java, Python, JPython, Ruby, JRuby, JavaScript (Node.js), PHP, .NET, Django, Clojure, Spring, Scala, or REST (any language)

Cypher (native/preferred), Native Java APIs (special cases), Traverser API, REST, Blueprints, Gremlin

ACID 42 / Very High

DEX Java, C++, .NET Native Java, C# and C++ APIs, Blueprints, Gremlin

Consistency, durability and partial isolation and atomicity

5 / ?

Data - Graph Databases - Neo4j customers

Data - Graph Database - Neo4j - Partners

EndThanks you for your attention.