The NoSQL Way in Postgres

NoSQL way in PostgreSQL
Vibhor Kumar (Principal System Engineer)

Agenda
• Intro to JSON, HSTORE and PL/V8
• JSON History in Postgres
• JSON Data Types, Operators and Functions
• JSON, JSONB– when to use which one?
• JSONB and Node.JS – easy as pie
• NoSQL Performance in Postgres – fast as greased lightning
• Say ‘Yes’ to ‘Not only SQL’
• Useful resources
© 2014 EnterpriseDB Corporation. All rights reserved. 2

Let’s Ask Ourselves, Why NoSQL?
• Where did NoSQL come from?
− Where all cool tech stuff comes from – Internet companies
• Why did they make NoSQL?
− To support huge data volumes and evolving demands for ways
to work with new data types
• What does NoSQL accomplish?
− Enables you to work with new data types: email, mobile
interactions, machine data, social connections
− Enables you to work in new ways: incremental development
and continuous release
• Why did they have to build something new?
− There were limitations to most relational databases

NoSQL: Real-world Applications
• Emergency Management System
− High variability among data sources required high schema
flexibility
• Massively Open Online Course
− Massive read scalability, content integration, low latency
• Patient Data and Prescription Records
− Efficient write scalability
• Social Marketing Analytics
− Map reduce analytical approaches
Source: Gartner, A Tour of NoSQL in 8 Use Cases,
by Nick Heudecker and Merv Adrian, February 28, 2014

Postgres’ Response
• HSTORE
− Key-value pair
− Simple, fast and easy
− Postgres v 8.2 – pre-dates many NoSQL-only solutions
− Ideal for flat data structures that are sparsely populated
• JSON
− Hierarchical document model
− Introduced in Postgres 9.2, perfected in 9.3
• JSONB
− Binary version of JSON
− Faster, more operators and even more robust
− Postgres 9.4

Postgres: Key-value Store
• Supported since 2006, the HStore
contrib module enables storing
key/value pairs within a single
column
• Allows you to create a schema-less,
ACID compliant data store within
Postgres
• Create single HStore column and
include, for each row, only those keys
which pertain to the record
• Add attributes to a table and query
without advance planning
•Combines flexibility with ACID compliance

HSTORE Examples
• Create a table with HSTORE field
CREATE TABLE hstore_data (data HSTORE);
• Insert a record into hstore_data
INSERT INTO hstore_data (data) VALUES (’
"cost"=>"500",
"product"=>"iphone",
"provider"=>"apple"');
• Select data from hstore_data
SELECT data FROM hstore_data ;
------------------------------------------
"cost"=>"500”,"product"=>"iphone”,"provider"=>"Apple"
(1 row)

Postgres: Document Store
• JSON is the most popular
data-interchange format on the web
• Derived from the ECMAScript
Programming Language Standard
(European Computer Manufacturers
Association).
• Supported by virtually every
programming language
• New supporting technologies
continue to expand JSON’s utility
− PL/V8 JavaScript extension
− Node.js
• Postgres has a native JSON data type (v9.2) and a JSON parser and a
variety of JSON functions (v9.3)
• Postgres will have a JSONB data type with binary storage and indexing
(coming – v9.4)

JSON Examples
• Creating a table with a JSONB field
CREATE TABLE json_data (data JSONB);
• Simple JSON data element:
{"name": "Apple Phone", "type": "phone", "brand":
"ACME", "price": 200, "available": true,
"warranty_years": 1}
• Inserting this data element into the table json_data
INSERT INTO json_data (data) VALUES
(’ { "name": "Apple Phone",
"type": "phone",
"brand": "ACME",
"price": 200,
"available": true,
"warranty_years": 1
} ')

JSON Examples
• JSON data element with nesting:
{“full name”: “John Joseph Carl Salinger”,
“names”:
[
{"type": "firstname", “value”: ”John”},
{“type”: “middlename”, “value”: “Joseph”},
{“type”: “middlename”, “value”: “Carl”},
{“type”: “lastname”, “value”: “Salinger”}
]
}

A simple query for JSON data
SELECT DISTINCT
data->>'name' as products
FROM json_data;
products
------------------------------
Cable TV Basic Service
Package
AC3 Case Black
Phone Service Basic Plan
AC3 Phone
AC3 Case Green
Phone Service Family Plan
AC3 Case Red
AC7 Phone
This query does not
return JSON data – it
returns text values
associated with the
key ‘name’

A query that returns JSON data
SELECT data FROM json_data;
data
------------------------------------------
{"name": "Apple Phone", "type": "phone",
"brand": "ACME", "price": 200,
"available": true, "warranty_years": 1}
This query returns the JSON data in its
original format

JSON and ANSI SQL - PB&J for the DBA
• JSON is naturally
integrated with ANSI SQL
in Postgres
• JSON and SQL queries
use the same language, the
same planner, and the same ACID compliant
transaction framework
• JSON and HSTORE are elegant and easy to use
extensions of the underlying object-relational model

JSON and ANSI SQL Example
SELECT DISTINCT
product_type,
data->>'brand' as Brand,
data->>'available' as Availability
FROM json_data
JOIN products
ON (products.product_type=json_data.data->>'name')
WHERE json_data.data->>'available'=true;
product_type | brand | availability
---------------------------+-----------+--------------
AC3 Phone | ACME | true
ANSI SQL
JSON
No need for programmatic logic to combine SQL and
NoSQL in the application – Postgres does it all

Bridging between SQL and JSON
Simple ANSI SQL Table Definition
CREATE TABLE products (id integer, product_name text );
Select query returning standard data set
SELECT * FROM products;
id | product_name
----+--------------
1 | iPhone
2 | Samsung
3 | Nokia
Select query returning the same result as a JSON data set
SELECT ROW_TO_JSON(products) FROM products;
{"id":1,"product_name":"iPhone"}
{"id":2,"product_name":"Samsung"}
{"id":3,"product_name":"Nokia”}

JSON Data Types
• 1. Number:
− Signed decimal number that may contain a fractional part and may use exponential
notation.
− No distinction between integer and floating-point
• 2. String
− A sequence of zero or more Unicode characters.
− Strings are delimited with double-quotation mark
− Supports a backslash escaping syntax.
• 3. Boolean
− Either of the values true or false.
• 4. Array
− An ordered list of zero or more values,
− Each values may be of any type.
− Arrays use square bracket notation with elements being comma-separated.
• 5. Object
− An unordered associative array (name/value pairs).
− Objects are delimited with curly brackets
− Commas to separate each pair
− Each pair the colon ':' character separates the key or name from its value.
− All keys must be strings and should be distinct from each other within that object.
• 6. null
− An empty value, using the word null
JSON is defined per RFC – 7159
For more detail please refer
http://tools.ietf.org/html/rfc7159

JSON Data Type Example
{
"firstName": "John", -- String Type
"lastName": "Smith", -- String Type
"isAlive": true, -- Boolean Type
"age": 25, -- Number Type
"height_cm": 167.6, -- Number Type
"address": { -- Object Type
"streetAddress": "21 2nd Street”,
"city": "New York”,
"state": "NY”,
"postalCode": "10021-3100”
},
"phoneNumbers": [ // Object Array
{ // Object
"type": "home”,
"number": "212 555-1234”
},
{
"type": "office”,
"number": "646 555-4567”
}
],
"children": [],
"spouse": null // Null
}

JSON 9.4 – New Operators and Functions
• JSON
− New JSON creation functions (json_build_object, json_build_array)
− json_typeof – returns text data type (‘number’, ‘boolean’, …)
• JSONB data type
− Canonical representation
− Whitespace and punctuation dissolved away
− Only one value per object key is kept
− Last insert wins
− Key order determined by length, then bytewise comparison
− Equality, containment and key/element presence tests
− New JSONB creation functions
− Smaller, faster GIN indexes
− jsonb subdocument indexes
− Use “get” operators to construct expression indexes on subdocument:
− CREATE INDEX author_index ON books USING GIN ((jsondata ->
'authors'));
− SELECT * FROM books WHERE jsondata -> 'authors' ? 'Carl
Bernstein'

JSON and BSON
• BSON – stands for
‘Binary JSON’
• BSON != JSONB
− BSON cannot represent an integer or
floating-point number with more than
64 bits of precision.
− JSONB can represent arbitrary JSON values.
• Caveat Emptor!
− This limitation will not be obvious during early
stages of a project!

JSON, JSONB or HSTORE?
• JSON/JSONB is more versatile than HSTORE
• HSTORE provides more structure
• JSON or JSONB?
− if you need any of the following, use JSON
− Storage of validated json, without processing or indexing it
− Preservation of white space in json text
− Preservation of object key order Preservation of duplicate object
keys
− Maximum input/output speed
• For any other case, use JSONB

JSONB and Node.js - Easy as π
• Simple Demo of Node.js to Postgres cnnection

JSON Performance Evaluation
• Goal
− Help our customers understand when to chose
Postgres and when to chose a specialty
solution
− Help us understand where the NoSQL limits of
Postgres are
• Setup
− Compare Postgres 9.4 to Mongo 2.6
− Single instance setup on AWS M3.2XLARGE
(32GB)
• Test Focus
− Data ingestion (bulk and individual)
− Data retrieval

Performance Evaluation
Generate 50 Million
JSON Documents
Load into MongoDB 2.6
(IMPORT)
Load into
Postgres 9.4
(COPY)
50 Million individual
INSERT commands
50 Million individual
INSERT commands
Multiple SELECT
statements
Multiple SELECT
statements
T1
T2
T3

NoSQL Performance Evaluation
Correction to earlier versions:
MongoDB console does not allow for
INSERT of documents > 4K. This
lead to truncation of the MongoDB
size by approx. 25% of all records in
the benchmark.

Performance Evaluations – Next Steps
• Initial tests confirm that Postgres’ can handle many
NoSQL workloads
• EDB is making the test scripts publically available
• EDB encourages community participation to
better define where Postgres should be used
and where specialty solutions are appropriate
• Download the source at
https://github.com/EnterpriseDB/pg_nosql_benchmark
• Join us to discuss the findings at
http://bit.ly/EDB-NoSQL-Postgres-Benchmark

Structured or Unstructured?
“No SQL Only” or “Not Only SQL”?
• Structures and standards emerge!
• Data has references (products link to catalogues;
products have bills of material; components appear in
multiple products; storage locations link to ISO country
tables)
• When the database has duplicate data entries, then the
application has to manage updates in multiple places –
what happens when there is no ACID transactional
model?

Ultimate Flexibility with Postgres

Say yes to ‘Not only SQL’
• Postgres overcomes many of the standard objections
“It can’t be done with a conventional database system”
• Postgres
− Combines structured data and unstructured data (ANSI SQL
and JSON/HSTORE)
− Is faster (for many workloads) than than the leading NoSQL-only
solution
− Integrates easily with Web 2.0 application development
environments
− Can be deployed on-premise or in the cloud
Do more with Postgres – the Enterprise NoSQL Solution

Useful Resources
• Postgres NoSQL Training Events
− Bruce Momjian & Vibhor Kumar @ pgEurope
− Madrid (Oct 21): Maximizing Results with JSONB and PostgreSQL
• Whitepapers @ http://www.enterprisedb.com/nosql-for-enterprise
− PostgreSQL Advances to Meet NoSQL Challenges (business
oriented)
− Using the NoSQL Capabilities in Postgres (full of code examples)
• Run the NoSQL benchmark
− https://github.com/EnterpriseDB/pg_nosql_benchmark

The NoSQL Way in Postgres

More Related Content

What's hot (20)

Viewers also liked (20)

Similar to The NoSQL Way in Postgres (20)

More from EDB (20)

Recently uploaded (20)

The NoSQL Way in Postgres

Editor's Notes