Non-Relational Databases

NoSQL, NewSQL, CAP and other acronyms

Created by Christian Köberl, May 2013

Intro

1960s

Hierarchical and network databases

1966 - IBM IMS

1970s

1970 - Edgar F. Codd: A Relational Model of Data for Large Shared Data Banks

1977 - IBM System R -> SQL

1979 - Oracle

1980s

Mid 80s: Rise of the RDMS and SQL

1986 - BerkeleyDB - key/value

1989 - Lotus Notes/Domino - document oriented

2000 and Beyond

Rise of NoSQL

2000 - Neo4j - graph database

2003 - Memcached - most used key/value cache

2005 - CouchDB

2006 - Google BigTable paper

Current Trends

Cloud

Big Data

Connectivity

Variety

Relational Databases

Why Relational?

Structured data
Separation of application and data
Multiple application using the same database
Standard (SQL)
Tools
People (education)

A tomicity

C onsistency

I solation

D urability

Transaction Isolation

	Dirty	Non-repeatable reads	Phantom reads
Read uncommited	may occur	may occur	may occur
Read commited	-	may occur	may occur
Repeatable read	-	-	may occur
Serializable	-	-	-

MVVC

Problems With Relational Databases

Custom Fields

Customer A

Customer B

Tags, Network Structure

Tags - Relational Solution (1)

Tags - Relational Solution (2)

Tags - Relational Solution (3)

Query For Things With Tag 'Java'

SELECT * 
  FROM Thing
  JOIN Thing_Tag 
    ON Thing.Id = Thing_Tag.Thing_Id
  JOIN Tag 
    ON Tag.Id = Thing_Tag.Tag_Id
 WHERE Tag.Tag = 'Java'

Query For Tags 'Java' And 'SQL'

SELECT * 
  FROM Thing
  JOIN Thing_Tag tt1
    ON Thing.Id = tt1.Thing_Id
  JOIN Thing_Tag tt2
    ON Thing.Id = tt2.Thing_Id
  JOIN Tag tag1
    ON tag1.Id = tt1.Tag_Id
  JOIN Tag tag2
    ON tag2.Id = tt2.Tag_Id
 WHERE tt1.Tag = 'Java' AND tt2.Tag = 'SQL'

Query For Tags 'Java' Or 'SQL'

SELECT * 
  FROM Thing
  JOIN Thing_Tag 
    ON Thing.Id = Thing_Tag.Thing_Id
  JOIN Tag 
    ON Tag.Id = Thing_Tag.Tag_Id
 WHERE Tag.Tag = 'Java' OR Tag.Tag = 'SQL'

+ eliminate duplicate entries

Query For Tag 'Java' But Not 'SQL'

SELECT * 
  FROM Thing
  JOIN Thing_Tag 
    ON Thing.Id = Thing_Tag.Thing_Id
  JOIN Tag 
    ON Tag.Id = Thing_Tag.Tag_Id
 WHERE Tag.Tag = 'Java'
   AND NOT EXISTS 
      (SELECT 1 
         FROM Thing_Tag
         JOIN Tag
           ON Tag.Id = Thing_Tag.Tag_Id
        WHERE Thing_Tag.Thing_Id = Thing.Id
          AND Tag.Tag = 'SQL')

Same Queries With Lucene (Solr)

Tag 'Java':
tag:Java
Tag 'Java' and 'SQL':
tag:Java AND tag:SQL
Tag 'Java' or 'SQL':
tag:Java OR tag:SQL
Tag 'Java' but not 'SQL':
tag:Java -tag:SQL

CAP

or the problem with ACID

CAP Theorem

It is impossible in the asynchronous network model to implement a read/write data object that guarantees the following properties:

Availability

Atomic consistency

in all fair executions (including those in which messages are lost).

Consistency

all nodes see the same data at the same time

Availability

guarantee that every request receives a response about whether it was successful or failed

Partition Tolerance

the system continues to operate despite arbitrary message loss or failure of part of the system