Introduction to Analytics

and Big Data - Hadoop

Rob Peglar
EMC Isilon
SNIA Legal Notice
The material contained in this tutorial is copyrighted by the SNIA.
Member companies and individual members may use this material in
presentations and literature under the following conditions:
Any slide or slides used must be reproduced in their entirety without
modification
The SNIA must be acknowledged as the source of any material used in the
body of any document containing material from these presentations.
This presentation is a project of the SNIA Education Committee.
Neither the author nor the presenter is an attorney and nothing in this
presentation is intended to be, or should be construed as legal advice or an
opinion of counsel. If you need legal advice or a legal opinion please
contact your attorney.
The information presented herein represents the author's personal opinion
and current understanding of the relevant issues involved. The author, the
presenter, and the SNIA do not assume any responsibility or liability for
damages arising out of any reliance on or use of this information.
NO WARRANTIES, EXPRESS OR IMPLIED. USE AT YOUR OWN RISK.
Introduction to Analytics and Big Data – Hadoop
© 2012 Storage Networking Industry Association. All Rights Reserved.
2
BIG DATA AND HADOOP

Data Challenges
Why Hadoop

Introduction to Analytics and Big Data – Hadoop

© 2012 Storage Networking Industry Association. All Rights Reserved.
Customer Challenges: The Data Deluge

IN 2010 THE DIGITAL UNIVERSE WAS

1.2 ZETTABYTES
IN A DECADE THE DIGITAL UNIVERSE WILL BE
35 ZETTABYTES

90% OF THE DIGITAL UNIVERSE IS

UNSTRUCTURED
IN 2011 THE DIGITAL UNIVERSE IS
300 QUADRILLION FILES
The Economist, Feb 25, 2010

Introduction to Analytics and Big Data – Hadoop

© 2012 Storage Networking Industry Association. All Rights Reserved.
Big Data Is Different than Business Intelligence

“TRADITIONAL BI”

Repetitive
“BIG DATA ANALYTICS”

Experimental, Ad Hoc Structured

Mostly Semi-Structured Operational

External + Operational GBs to 10s of TBs

10s of TB to 100’s of PB’s

Introduction to Analytics and Big Data – Hadoop

© 2012 Storage Networking Industry Association. All Rights Reserved.
Questions from Businesses will Vary

Past Future

What What is What is likely to

happened? happening? happen?

Reporting, Real-Time Predictive

Dashboards Analytics Analytics

Why did it Why is it What should I do

happen? happening? about it?
Forensics & Data Real-Time Prescriptive
Mining Data Mining Analytics

Introduction to Analytics and Big Data – Hadoop

© 2012 Storage Networking Industry Association. All Rights Reserved.
Web 2.0 is “Data-Driven”

“The future is here, it’s just not evenly distributed yet.”

William Gibson

Introduction to Analytics and Big Data – Hadoop

© 2012 Storage Networking Industry Association. All Rights Reserved.
The world of Data-Driven Applications

Introduction to Analytics and Big Data – Hadoop

© 2012 Storage Networking Industry Association. All Rights Reserved.
Attributes of Big Data
Volume

Terabytes
Transactions
Tables
Records
Files

Batch
Structured
Near Time
Unstructured
Real Time
Semistructured
Streams

Velocity Variety
Introduction to Analytics and Big Data – Hadoop
© 2012 Storage Networking Industry Association. All Rights Reserved.
Ten Common Big Data Problems

1. Modeling true risk 6. Analyzing network

2. Customer churn data to predict
analysis failure
3. Recommendation 7. Threat analysis
engine 8. Trade surveillance
4. Ad targeting 9. Search quality
5. PoS transaction 10.Data “sandbox”
analysis

Introduction to Analytics and Big Data – Hadoop

© 2012 Storage Networking Industry Association. All Rights Reserved.
The Big Data Opportunity

Financial Services Healthcare

Retail Web/Social/Mobile

Manufacturing Government

Introduction to Analytics and Big Data – Hadoop

© 2012 Storage Networking Industry Association. All Rights Reserved.
Industries Are Embracing Big Data

Retail Advertising & Public Relations

• CRM – Customer Scoring • Demand Signaling
• Store Siting and Layout • Ad Targeting
• Fraud Detection / Prevention • Sentiment Analysis
• Supply Chain Optimization • Customer Acquisition

Financial Services Media & Telecommunications

• Algorithmic Trading • Network Optimization
• Risk Analysis • Customer Scoring
• Fraud Detection • Churn Prevention
• Portfolio Analysis • Fraud Prevention

Manufacturing Energy
• Product Research • Smart Grid
• Engineering Analytics • Exploration
• Process & Quality Analysis
• Distribution Optimization

Government Healthcare & Life Sciences

• Market Governance • Pharmaco-Genomics
• Counter-Terrorism • Bio-Informatics
• Econometrics • Pharmaceutical Research
• Health Informatics • Clinical Outcomes Research

Introduction to Analytics and Big Data – Hadoop

© 2012 Storage Networking Industry Association. All Rights Reserved.
Why Hadoop?

Answer: Big Datasets!

Introduction to Analytics and Big Data – Hadoop

© 2012 Storage Networking Industry Association. All Rights Reserved.
Why Hadoop?

Big Data analytics and the Apache Hadoop open source

project are rapidly emerging as the preferred solution to
address business and technology trends that are
disrupting traditional data management and processing.
Enterprises can gain a competitive advantage by
being early adopters of big data analytics.

Introduction to Analytics and Big Data – Hadoop

© 2012 Storage Networking Industry Association. All Rights Reserved.
Storage & Memory B/W lagging CPU

CPU DRAM LAN Disk

Annual bandwidth improvement (all milestones)
1.5 1.27 1.39 1.28

Annual latency improvement (all milestones) 1.17 1.07 1.12 1.11

Memory Wall Storage Chasm

CPU B/W requirements out-pacing memory and

storage
Disk & memory getting “further” away from CPU
Large sequential transfers better for both memory &
disk

Introduction to Analytics and Big Data – Hadoop

© 2012 Storage Networking Industry Association. All Rights Reserved.
Commodity Hardware Economics

For $1000 Process

One computer can ~32GB

Store 99.9%
~15TB Of data is Underutilized

Introduction to Analytics and Big Data – Hadoop

© 2012 Storage Networking Industry Association. All Rights Reserved.
Enterprise + Big Data = Big Opportunity

Introduction to Analytics and Big Data – Hadoop

© 2012 Storage Networking Industry Association. All Rights Reserved.
17
WHAT IS HADOOP
Hadoop Adoption
HDFS
MapReduce
Ecosystem Projects

Introduction to Analytics and Big Data – Hadoop

© 2012 Storage Networking Industry Association. All Rights Reserved.
Hadoop Adoption in the Industry
2007 2008 2009 2010

The Datagraph Blog

Source:
Introduction Hadoopand
to Analytics Summit
Big DataPresentations
– Hadoop
© 2012 Storage Networking Industry Association. All Rights Reserved.
What is Hadoop?

A scalable fault-tolerant distributed system for data storage and

processing
Core Hadoop has two main components
Hadoop Distributed File System (HDFS): self-healing, high-bandwidth clustered
storage
Reliable, redundant, distributed file system optimized for large files
MapReduce: fault-tolerant distributed processing
Programming model for processing sets of data
Mapping inputs to outputs and reducing the output of multiple Mappers to
one (or a few) answer(s)
Operates on unstructured and structured data
A large and active ecosystem
Open source under the friendly Apache License
http://wiki.apache.org/hadoop/

Introduction to Analytics and Big Data – Hadoop

© 2012 Storage Networking Industry Association. All Rights Reserved.
HDFS 101
The Data Set System

Introduction to Analytics and Big Data – Hadoop

© 2012 Storage Networking Industry Association. All Rights Reserved.
HDFS Concepts

 Sits on top of a native (ext3, xfs, etc..) file system

 Performs best with a ‘modest’ number of large files
 Files in HDFS are ‘write once’
 HDFS is optimized for large, streaming reads of files

Introduction to Analytics and Big Data – Hadoop

© 2012 Storage Networking Industry Association. All Rights Reserved.
HDFS
 Hadoop Distributed File System
– Data is organized into files & directories
– Files are divided into blocks, distributed across
cluster nodes
– Block placement known at runtime by map-
reduce = computation co-located with data
– Blocks replicated to handle failure
– Checksums used to ensure data integrity
 Replication: one and only strategy for error
handling, recovery and fault tolerance
– Self Healing
– Make multiple copies

Introduction to Analytics and Big Data – Hadoop

© 2012 Storage Networking Industry Association. All Rights Reserved.
Hadoop Server Roles

Client Client Client Client Client Client Client Client

Name Job Secondary

Node Tracker Node

Master Master

Data Task Data Data Task

Task Tracker
Node Tracker Node Node Tracker

Slave Slave Slave

Up to 4K
Nodes
Data Task Data Data
Task Tracker Task Tracker
Node Tracker Node Node

Slave Slave Slave

Introduction to Analytics and Big Data – Hadoop

© 2012 Storage Networking Industry Association. All Rights Reserved.
Hadoop Cluster

CORE SWITCH CORE SWITCH Client

1GbE/10GbE 1GbE/10GbE 1GbE/10GbE 1GbE/10GbE

NN JT SNN DN, TT

DN, TT DN, TT DN, TT DN, TT

DN, TT DN, TT DN, TT DN, TT

Up to 4K
DN, TT DN, TT DN, TT DN, TT
Nodes
DN, TT DN, TT DN, TT DN, TT

DN, TT DN, TT DN, TT DN, TT

DN, TT DN, TT DN, TT DN, TT

Introduction to Analytics and Big Data – Hadoop

© 2012 Storage Networking Industry Association. All Rights Reserved.
HDFS File Write Operation

Introduction to Analytics and Big Data – Hadoop

© 2012 Storage Networking Industry Association. All Rights Reserved.
HDFS File Read Operation

Introduction to Analytics and Big Data – Hadoop

© 2012 Storage Networking Industry Association. All Rights Reserved.
MapReduce 101
Functional Programming meets
Distributed Processing

Introduction to Analytics and Big Data – Hadoop

© 2012 Storage Networking Industry Association. All Rights Reserved.
MapReduce Provides

Automatic parallelization and distribution

Fault Tolerance
Status and Monitoring Tools
A clean abstraction for programmers
Google Technology RoundTable: MapReduce

Introduction to Analytics and Big Data – Hadoop

© 2012 Storage Networking Industry Association. All Rights Reserved.
What is MapReduce?

A method for distributing a task across multiple nodes

Each node processes data stored on that node
Consists of two developer-created phases
1. Map
2. Reduce
In between Map and Reduce is the Shuffle and Sort

Introduction to Analytics and Big Data – Hadoop

© 2012 Storage Networking Industry Association. All Rights Reserved.
Key MapReduce Terminology Concepts

A user runs a client program on a client computer

The client program submits a job to Hadoop
The job is sent to the JobTracker process on the
Master Node
Each Slave Node runs a process called the
TaskTracker
The JobTracker instructs TaskTrackers to run and
monitor tasks
A task attempt is an instance of a task running on a
slave node
There will be at least as many task attempts as there
are tasks which need to be performed
Introduction to Analytics and Big Data – Hadoop
© 2012 Storage Networking Industry Association. All Rights Reserved.
MapReduce: Basic Concepts

Each Mapper processes single input split from HDFS

Hadoop passes developer’s Map code one record at a
time
Each record has a key and a value
Intermediate data written by the Mapper to local disk
During shuffle and sort phase, all values associated
with same intermediate key are transferred to same
Reducer
Reducer is passed each key and a list of all its values
Output from Reducers is written to HDFS
Introduction to Analytics and Big Data – Hadoop
© 2012 Storage Networking Industry Association. All Rights Reserved.
MapReduce Operation

What was the max/min temperature for the last century?

Introduction to Analytics and Big Data – Hadoop

© 2012 Storage Networking Industry Association. All Rights Reserved.
Sample Dataset

The requirement:
you need to find out grouped by type of customer how
many of each type are in each country with the name of the
country listed in the countries.dat in the final result
(and not the 2 digit country name). Each record has a key
and a value
To do this you need to:
Join the data sets
Key on country
Count type of customer per country
Output the results

Introduction to Analytics and Big Data – Hadoop

© 2012 Storage Networking Industry Association. All Rights Reserved.
MapReduce Paradigm

Input Map Shuffle and Sort Reduce Output

Map

Reduce

Map

Reduce

Map

cat grep sort uniq output

Introduction to Analytics and Big Data – Hadoop

© 2012 Storage Networking Industry Association. All Rights Reserved.
 MapReduce Example
Problem: Count the number of times that each word appears in the following paragraph:
John has a red car, which has no radio. Mary has a red
bicycle. Bill has no car or bicycle.

Server 1: John has a red car, which has no radio. Server 2: Mary has a red bicycle. Server 3: Bill has no car or bicycle.
John: 1 Mary: 1 Bill: 1
has: 2 has: 1 has: 1
a: 1 a: 1 no: 1
Map red: 1 red: 1 car: 1
car: 1 bicycle: 1 or: 1
which: 1 biclycle:1
no: 1
radio: 1

Server 1 Server 2 Server 3 Server 1 Server 2 Server 3

John: car: 1 bicycle: 1 John: 1 car: 2 bicycle: 2
1 car: 1 bicycle: 1 has 4 which: 1 Bill: 1
has 2 which: 1 Bill: 1 a: 2 no: 2 or: 1
has: 1
Reduce has: 1
no: 1
no: 1
or: 1 red: 2 radio: 1
Mary: 1
a: 1 radio: 1
a: 1 Mary: 1
red: 1
red: 1

Introduction to Analytics and Big Data – Hadoop

© 2012 Storage Networking Industry Association. All Rights Reserved.
 Putting it all Together:
MapReduce and HDFS
Job Tracker
Client/Dev
Map Job
2
Map
Map Job
Job Reduce Job

Reduce
Reduce Job
Job
3
Task Tracker
Task Tracker
Task Tracker

Map Job 4 Map Job Map Job

Reduce Job Reduce Job Reduce Job

1

Large Data Set

(Log files, Sensor Data) Hadoop Distributed File System (HDFS)

Introduction to Analytics and Big Data – Hadoop

© 2012 Storage Networking Industry Association. All Rights Reserved.
Hadoop Ecosystem Projects
• Hadoop is a ‘top-level’ Apache project
• Created and managed under the auspices of the Apache Software Foundation

• Several other projects exist that rely on some or all of Hadoop

• Typically either both HDFS and MapReduce, or just HDFS

• Ecosystem Projects Include

• Hive
• Pig
• HBase
• Many more…..

http://hadoop.apache.org/
Introduction to Analytics and Big Data – Hadoop
© 2012 Storage Networking Industry Association. All Rights Reserved.
Hadoop, SQL & MPP Systems

Hadoop Traditional SQL MPP Systems

Systems
Scale-Out Scale-Up Scale-Out

Key/Value Pairs Relational Tables Relational Tables

Functional Declarative Queries Declarative Queries

Programming
Offline Batch Online Transactions Online Transactions
Processing

Introduction to Analytics and Big Data – Hadoop

© 2012 Storage Networking Industry Association. All Rights Reserved.
Comparing RDBMS and MapReduce

Traditional RDBMS MapReduce

Data Size Gigabytes (Terabytes) Petabytes (Exabytes)

Access Interactive and Batch Batch

Updates Read / Write many times Write once, Read many times

Structure Static Schema Dynamic Schema

Integrity High (ACID) Low

Scaling Nonlinear Linear

DBA Ratio 1:40 1:3000

Reference: Tom White’s Hadoop: The Definitive Guide

Introduction to Analytics and Big Data – Hadoop

© 2012 Storage Networking Industry Association. All Rights Reserved.
Hadoop Use Cases

Introduction to Analytics and Big Data – Hadoop

© 2012 Storage Networking Industry Association. All Rights Reserved.
Diagnostics and Customer Churn

Issues
What make and model systems are deployed?
Are certain set top boxes in need of replacement based on system
diagnostic data?
Is the a correlation between make, model or vintage of set top box and
customer churn?
What are the most expensive boxes to maintain?
Which systems should we pro-actively replace to keep customers happy?
Big Data Solution
Collect unstructured data from set top boxes—multiple terabytes
Analyze system data in Hadoop in near real time
Pull data in to Hive for interactive query and modeling
Analytics with Hadoop increases customer satisfaction

Introduction to Analytics and Big Data – Hadoop

© 2012 Storage Networking Industry Association. All Rights Reserved.
Pay Per View Advertising
Issues
Fixed inventory of ad space is provided by national content providers. For
example, 100 ads offered to provider for 1 month of programming
Provider can use this space to advertise its products and services, such as
pay per view
Do we advertise “The Longest Yard” in the middle of a football game or in
the middle of a romantic comedy?
10% increase in pay per view movie rentals = $10M in incremental revenue
• Big Data Solution
Collect programming data and viewer rental data in a large data repository
Develop models to correlate proclivity to rent to programming format
Find the most productive time slots and programs to advertise pay per
view inventory
Improve ad placement and pay-per-view conversion with Hadoop

Introduction to Analytics and Big Data – Hadoop

© 2012 Storage Networking Industry Association. All Rights Reserved.
Risk Modeling
 Risk Modeling
– Bank had customer data across multiple lines of business and needed to
develop a better risk picture of its customers. i.e, if direct deposits stop
coming into checking acct, it’s likely that customer lost his/her job, which
impacts creditworthiness for other products (CC, mortgage, etc.)
– Data existing in silos across multiple LOB’s and acquired bank systems
– Data size approached 1 petabyte
 Why do this in Hadoop?
– Ability to cost-effectively integrate + 1 PB of data from multiple data
sources: data warehouse, call center, chat and email
– Platform for more analysis with poly-structured data sources; i.e.,
combining bank data with credit bureau data; Twitter, etc.
– Offload intensive computation from DW

Introduction to Analytics and Big Data – Hadoop

© 2012 Storage Networking Industry Association. All Rights Reserved.
Sentiment Analysis

 Sentiment Analysis
– Hadoop used frequently to monitor what customers think of
company’s products or services
– Data loaded from social media sources (Twitter, blogs,
Facebook, emails, chats, etc.) into Hadoop cluster
– Map/Reduce jobs run continuously to identify sentiment (i.e.,
Acme Company’s rates are “outrageous” or “rip off”)
– Negative/positive comments can be acted upon (special offer,
coupon, etc.)
 Why Hadoop
– Social media/web data is unstructured
– Amount of data is immense
– New data sources arise weekly
Introduction to Analytics and Big Data – Hadoop
© 2012 Storage Networking Industry Association. All Rights Reserved.
Resources to enable the Big Data Conversation

World Economic Forum: “Personal Data: The Emergence of a New Asset

Class” 2011
McKinsey Global Institute: Big Data: The next frontier for innovation,
competition, and productivity
Big Data: Harnessing a game-changing asset
IDC: 2011 Digital Universe Study: Extracting Value from Chaos
The Economist: Data, Data Everywhere
Data Science Revealed: A Data-Driven Glimpse into the Burgeoning New
Field
O’Reilly – What is Data Science?
O’Reilly – Building Data Science Teams?
O’Reilly – Data for the public good
Obama Administration “Big Data Research and Development Initiative.”

Introduction to Analytics and Big Data – Hadoop

© 2012 Storage Networking Industry Association. All Rights Reserved.
Q&A / Feedback

Please send any questions or comments on this

presentation to the SNIA at this address:
[email protected]
Many thanks to the following individuals
for their contributions to this tutorial.
SNIA Education Committee

Denis Guyadeen
Rob Peglar

Introduction to Analytics and Big Data – Hadoop

© 2012 Storage Networking Industry Association. All Rights Reserved.
47