Why no sql

Concept introduction series
Anand, 26-Feb-2015

What this deck aims to convey and what it doesn’t?

3
Captures some attributes of
RDBMS (SQL DB!) with the
intent of providing the context
behind non-relational database
models
Does not attempt to capture
the pro-s or con-s of RDBMS.
Does not attempt to explain
RDBMS’ role in modern server
architectures etc.

Some attributes and their implications

5
{
id: 223091
job: “test.feature_x”
status: “complete”
events:
{
{id: 220923232, type: started, time: 13:20:21, date: 03-06-2016}
{id: 220923439, type: restarted, time: 13:20:30, date: 03-06-2016}
{id: 220923439, type: finished, time: 12:24:31, date: 03-06-2016}
}
{description: {console_output: {blah blah}}
}

6
{
id: 223091
job: “test.feature_x”
status: “complete”
events:
{
{id: 220923232, type: started, time: 13:20:21, date: 03-06-2016}
{id: 220923439, type: restarted, time: 13:20:30, date: 03-06-2016}
{id: 220923439, type: finished, time: 12:24:31, date: 03-06-2016}
}
{description: {console_output: {blah blah}}
}
Jobs
Job Definition
Events
Status

7
An event in job table must already have an entry in event table (create an entry in event table and then
attach it to a job table to which it belongs). Schema changes (version) management is not trivial either.

8
Thinking in tables needs a different skillset than creating a Job class. In other words, memory data
structures (classes/objects in your heap or in a in-mem db) rarely resemble relational tables. Keeping
them in-synch while rapid development/ prototyping is costly.
Jobs
Job Definition
Events
Status
=

9
Clustering
(sharding,
replicating etc.)
data is a necessity
with constantly
evolving
relationships
(=schema) and
exponentially
increasing
volumes. Doing
that in RDBMS is
complicated to say
the least

11
SELECT db_name_text
FROM db_table
WHERE data_relationship_model <>
‘TABULAR’;

12
KEY VALUE
{
User Id: <id>
User:
{
Call Name: X
Address: Y
Genres: [“action”, “noir”]
}
}
{
User Id: <id>
User:
{
Call Name: X
Address: Y
}
}
{
User Id: <id>
User:
{
Call Name: X
Address: Y
}
}
DOCUMENT
GRAPH
{
User Id: <id>
User:
{
Call Name: X
Address: Y
}
}
{
User Id: <id>
User:
{
Call Name: X
Address: Y
}
}
{
Event Id: <id>
Event:
{
Lat: 12.952772
Long: 77.644305
Type: <location alert>
}
}
COLUMN FAMILY
a
c
b
x
z
y

13
= A Key to a value mapping storage
Think of java.util.Map<K,V>
 The DB knows only about the keys
 Values could be anything; they are just stored and retrieved,
never processed
Use cases:
– User -> Profile Data
– User -> Preference Data
– User -> Session Data

{
User Id: <id>
User:
{
Call Name: X
Address: Y
}
}
{
User Id: <id>
User:
{
Call Name: X
Address: Y
}
}
{
User Id: <id>
User:
{
Call Name: X
Address: Y
}
}
{
User Id: <id>
User:
{
Call Name: X
Address: Y
}
}
{
User Id: <id>
User:
{
Call Name: X
Address: Y
}
}
{
Event Id: <id>
Event:
{
Lat: 12.952772
Long: 77.644305
Type: <location alert>
}
}
= A (implicit) Key (id) to a document (value) mapping storage
Think of filing a document in a folder
 Values are supplied in readable formats – json, bson, XML
 Values are visible to the database and typically indexed
Use cases:
– Rapidly evolving data (application of schema like thing can
be delayed till maturity)
– Event processing (self contained documents for future
computation; mostly ‘read’ or ‘created’ and rarely ‘updated’)

= Column is the unit of data; Stores columns as groups (‘families’)
Think of it as a ‘sparse matrix’
 Every row (key) may have data only against a sub-set of
columns
Use cases:
– Live analytics, Iteratively computed outcomes
{column family
‘likes’
user1
user2
userX
userY

= Graphs with data in both nodes and edges
Think of it as a, well, just graph!
 Node data represent what we normally think as ‘data’
 Edge data represent the ‘relationship’ between the connected
nodes
Use cases:
– Recommendations, Relationship traversal & analytics
a
c
b
x
z
y

18
 No SQL Distilled
 The Definitive Guide to MongoDB
 Plenty of references in WWW

19
 DB Schema: http://commons.wikimedia.org/wiki/File:Mediawiki-database-schema.png
 Internet per Minute: http://upload.wikimedia.org/wikipedia/commons/5/54/Internet_Minute_Infographic.jpg
 NoSQL Boom: From Ramana under CC2.0

Why no sql

More Related Content

Why no sql