Hadoop Master Class : A concise overview

Big Data Hadoop Master Class
By Abhishek Roy

About Me
•
•
•
•
•

Abhishek Roy
Around 10 years of tech experience with services,
product and startups
Was leading the data team at Qyuki Digital Media
Lately involved in Big data with emphasis on
recommendation systems and machine learning
approaches
Currently working on building an employee wellness
platform, www.feetapart.com
Linkedin : http://www.linkedin.com/in/abhishekroy8

Agenda
•
•
•
•
•
•
•
•

What is Big Data
History and Background
Why Data
Intro to Hadoop
Intro to HDFS
Setup & some hands on
HDFS architecture
Intro to Map Reduce

Agenda
• Pig
• Hive
• Apache Mahout
– Running an example MR Job

• Sqoop, Flume, Hue, Zookeeper
• Impala
• Some Real World Use cases

What Is Big Data?
Like the term Cloud, it is a bit illdefined/hazy

What’s the “BIG” hype about Big Data?
There may be hype, but problems are real and of big
value. How?
• We are in the age of advanced analytics (that’s
where all the problem is ,we want to analyze the
data) where valuable business insight is mined out
of historical data.
• We live in the age of crazy data where every
individual , enterprise and machine leaves so much
data behind summing up to many terabytes &
petabytes , and it is only expected to grow.

What’s the “BIG” hype about Big Data?
• Good news. Blessing in disguise. More data
means more precision.
• More data usually beats better algorithms.
• But how are we going to analyze?
• Traditional database or warehouse systems
crawl or crack at these volumes.
• Inflexible to handle most of these formats.
• This is the very characteristic of Big Data.

Key Hadoop/Big Data Data Sources
•
•
•
•
•
•
•
•
•

Sentiment
Clickstream
Sensor/Machine
Geographic
Text
Chatter from social networks
Traffic flow sensors
Satellite imagery
Broadcast audio streams

Sources of Big Data
•
•
•
•
•
•
•

Banking transactions
MP3s of rock music
Scans of government documents
GPS trails
Telemetry from automobiles
Financial market data
….

Key Drivers
Spread of cloud computing, mobile
computing and social media
technologies, financial transactions

Introduction cont.
• Nature of Big Data
– Huge volumes of data that cannot be handled by
traditional database or warehouse systems.
– It is mostly machine produced.
– Mostly unstructured and grows at high velocity.
– Big data doesn’t always mean huge data, it means
“difficult” data.

Inflection Points
• Is divide the data and rule a solution here?
– Have a multiple disk drive , split your data file into
small enough pieces across the drives and do
parallel reads and processing.
– Hardware Reliability (Failure of any drive) is a
challenge.
– Resolving data interdependency between drives is
a notorious challenge.
– Number of disk drives that can be added to a
server is limited

Inflection Points
• Analysis
– Much of big data is unstructured. Traditional
RDBMS/EDW cannot handle it.
– Lots of Big Data analysis is adhoc in nature, involves
whole data scan, referencing itself, joining, combing
etc.
– Traditional RDBMS/EDW cannot handle these with
their limited scalability options and architectural
limitations.
– You can incorporate betters servers, processors and
throw in more RAM but there is a limit to it.

Inflection Points
• We need a Drastically different approach
– A distributed file system with high capacity and
high reliability.
– A process engine that can handle structure
/unstructured data.
– A computation model that can operate on
distributed data and abstracts data dispersion.

What is Hadoop?
• “framework for running [distributed]
applications on large cluster built of
commodity hardware” –from Hadoop Wiki
• Originally created by Doug Cutting
– Named the project after his son’s toy

• The name “Hadoop” has now evolved to cover
a family of products, but at its core, it’s
essentially just the MapReduce programming
paradigm + a distributed file system

Pig
A platform for analyzing large data sets that
consists of a high-level language for
expressing data analysis programs, coupled
with infrastructure for evaluating these
programs.

Mahout
A machine learning library with algorithms
for clustering, classification and batch
based collaborative filtering that are
implemented on top of Apache Hadoop.

Hive
Data warehouse software built on top of
Apache Hadoop that facilitates querying
and managing large datasets residing in
distributed storage.

Sqoop
A tool designed for efficiently transferring
bulk data between Apache Hadoop and
structured data stores such as relational
databases.

Apache Flume
A distributed service for collecting,
aggregating, and moving large log data
amounts to HDFS.

Twitter Storm
Storm can be used to process a
stream of new data and update
databases in real time.

Funding & IPO
• Cloudera, (Commerical Hadoop) more than
$75 million
• MapR (Cloudera competitor) has raised more
than $25 million
• 10Gen (Maker of the MongoDB) $32 million
• DataStax (Products based on Apache
Cassandra) $11 million
• Splunk raised about $230 million through IPO

Big Data Application Domains
•
•
•
•
•
•

Healthcare
The public sector
Retail
Manufacturing
Personal-location data
Finance

Use The Right Tool For The Right Job
Relational Databases:

When to use?

Hadoop:

When to use?

•

Interactive Reporting (<1sec)

•

Affordable Storage/Compute

•

Multistep Transactions

•

Structured or Not (Agility)

•

Lots of Inserts/Updates/Deletes

•

Resilient Auto Scalability

Ship the Function to the Data
Traditional Architecture
function

Distributed Computing
function

data

data

data

data

function

function

function

data

data

function

data

RDBMS

function

data

data

data

data

data

data

data

data

function

function

function

data

data

data

data

data

data

data

data

data

function

function

function

data

data

data

SAN/NAS

Economics of Hadoop Storage
• Typical Hardware:
–
–
–
–

Two Quad Core Nehalems
24GB RAM
12 * 1TB SATA disks (JBOD mode, no need for RAID)
1 Gigabit Ethernet card

• Cost/node: $5K/node
• Effective HDFS Space:
– ¼ reserved for temp shuffle space, which leaves 9TB/node
– 3 way replication leads to 3TB effective HDFS space/node
– But assuming 7x compression that becomes ~ 20TB/node

Effective Cost per user TB: $250/TB
Other solutions cost in the range of $5K to $100K per user
TB

Market and Market Segments
Research Data and Predictions

http://wikibon.org/wiki/v/Big_Data_Market_Size_and_Vendor_Revenues

Market for big data tools will rise
from $9 billion to $86 billion in 2020

Future of Big Data
• More Powerful and Expressive Tools for Analysis
• Streaming Data Processing (Storm from Twitter and S4 from
Yahoo)
• Rise of Data Market Places (InfoChimps, Azure
Marketplace)
• Development of Data Science Workflows and Tools (Chorus,
The Guardian, New York Times)
• Increased Understanding of Analysis and Visualization

http://www.evolven.com/blog/big-data-predictions.html

Skills Gap
•
•
•
•
•

Statistics
Operations Research
Math
Programming
So-called "Data Hacking"

Typical Hadoop Architecture
Business Users

End Customers

Business Intelligence
OLAP Data Mart

Interactive Application
OLTP Data Store

Engineers
Hadoop: Storage and Batch Processing

Data Collection

HDFS Introduction
• Written in Java
• Optimized for larger files
– Focus on streaming data
(high-throughput > low-latency)

• Rack-aware
• Only *nix for production env.
• Web consoles for stats

MapReduce basics
• Take a large problem and divide it into subproblems
• Perform the same function on all subproblems
• Combine the output from all sub-problems

MapReduce(M/R) facts
• M/R is excellent for problems where the “subproblems” are not interdependent
• For example, the output of one “mapper” should
not depend on the output or communication with
another “mapper”
• The reduce phase does not begin execution until
all mappers have finished
• Failed map and reduce tasks get auto-restarted
• Rack/HDFS-aware

What is the MapReduce Model
• MapReduce is a computation model that supports parallel
processing on distributed data using clusters of computers.
• The MapReduce model expects the input data to be split
and distributed to the machines on the clusters so that
each splits cab be processed independently and in parallel.
• There are two stages of processing in MapReduce model to
achieve the final result: Map and Reduce. Every machine
in the cluster can run independent map and reduce
processes.

What is MapReduce Model
• Map phase processes the input splits. The output of the Map phase
is distributed again to reduce processes to combine the map output
to give final expected result.
• The model treats data at every stage as a key and value pair,
transforming one set of key/value pairs into different set of
key/value pairs to arrive at the end result.
• Map process tranforms input key/values pairs to a set of
intermediate key/vaue pairs.

• MapReduce framework passes this output to reduce processes
which will transform this to get final result which again will be in the
form of key/value pairs.

Design of MapReduce -Daemons
The MapReduce system is managed by two daemons
• JobTracker & TaskTracker
• JobTracker TaskTracker function in master / slave fashion
– JobTracker coordinates the entire job execution
– TaskTracker runs the individual tasks of map and
reduce
– JobTracker does the bookkeeping of all the tasks run on
the cluster
– One map task is created for each input split
– Number of reduce tasks is configurable
(mapred.reduce.tasks)

Who Loves it
• Yahoo !! Runs 20,000 servers running Hadoop
• Largest Hadoop clusters is 4000 servers,16PB
raw storage
• Facebook runs 2000 Hadoop servers
• 24 PB raw storage and 100 TB raw log/day
• eBay and LinkedIn has production use of
Hadoop.
• Sears retails uses Hadoop.

Hadoop Requirements
• Supported Platforms
• GNU/Linux is supported as development and
production

• Required Software
•
•

java 1.6.x +
ssh to be installed ,sshd must be running (for machined in
the cluster to interact with master machines)

• Development Environment
• Eclipse 3.6 or above

Lab Requirements
• Window 7-64 bit OS , Min 4 GB Ram
• VMWare Player5.0.0
• Linux VM-Ubuntu 12.04 LTS
– User: hadoop, password :hadoop123
• Java 6 installed on Linux VM
• Open SSH installed on Linux VM
• Putty-For opening Telnet sessions to the Linux VM
• WinSCP- For transferring files between windows and
Linux VM
• Other Linux machines will do as well
• Eclipse 3.6

Extract VM
• Folder:
Hadoop_MasterClassUbuntu_VM
• Copy:
ubuntu-12.04.1-server-amd64_with_hadoop.zip
• Extract to local using 7-zip

Starting VM

• Enter userID/password
• Type ifconfig
– Note down the ip address
– Connect to the VM using Putty

Install and Configure ssh(non-VM
users)
• Install ssh
sudo apt –get install ssh
– Check ssh installation: which ssh
which sshd
which ssh-keygen

– Generate ssh Key
ssh-keygen –t rsa –P ” –f ~/.ssh/id_rsa

– Copy public key as an authorized key(equivalent to
slave node)
cp ~/.ssh/id_rsa.pub ~/.ssh/authorized_keys
chmod 700 ~/.ssh
chmod 600 ~/.ssh/authorized_keys

Install and Configure ssh
• Verify SSHby logging into target(localhost
here)
– Command:
ssh localhost
– (this command will log you into the machine
localhost)

Accessing VM Putty and WinSCP
• Get IP address of the VM by using command
in Linus VM
• Use Putty to telnet to VM
• Use WinSCP to FTP files to VM

Accesing VM using Putty & WinSCP
Putty

WinSCP

Lab-VM Directory Structure(non-VM
users)
• User Home directory for user “hadoop”(created default
by OS)
– /home/hadoop

• Create working directory for the lab session
– /home/hadoop/lab

• Create directory for storing all downloads(installables)
– /home/hadoop/lab/downloads

• Create directory for storing data for analysis
– /home/hadoop/lab/data

• Create directory for installing tools
– /home/hadoop/lab/install

Install and Configure Java(non-VM
users)
• Install Open JDK
– Command :
– sudo apt-get install openjdk-6-jdk
– Check installation :java -version

• Configure java home in environment
– Add a line to .bash_profile to set java Home
• export JAVA_HOME =/usr/lib/jvm/java-6-openjdkamd64

– Hadoop will use during runtime

Install Hadoop(non-VM users)
•

Download hadoop jar with wget
– http://archive.apache.org/dist/hadoop/core/hadoop-1.0.3/hadoop-1.0.3.tar.gz

•

Untar
– cd ~/lab/install
– tar xzf ~/lab/downloads/hadoop-1.03.tar.gz
– Check extracted directory “hadoop-1.0.3”

•

Configure environment in .bash_profile ( or .bashrc)
– Add below two lines and execute bash profile
• export HADOOP_INSTALL=~/lab/install/hadoop-1.0.3
• export PATH=$PATH:$HADOOP_INSTALL/bin
• . .bash_profile (Execute bashrc)

•

Check Hadoop installation
– hadoop version

Setting up Hadoop
• Open $HADOOP_HOME/conf/hadoop-env.sh
• set the JAVA_HOME environment variable to
the $JAVA_HOME directory.
export JAVA_HOME=/usr/lib/jvm/java-6-openjdkamd64

Component of core Hadoop
•

At a high-level Hadoop architecture components can be classified into two
categories
– Distributed File management system-HDFS

•

This has central and distributed sub components
•
•

DataNode: Take care of the local file segments and constantly communicates with NameNode

•

•
•

NameNode: Centrally Monitors and controls the whole file system

Secondary NameNode:Do not confuse. This is not a NameNode Backup. This just backs up the file
system status from the NameNode periodically.

Distributed computing system-MapReduce Framework
This again has central and distributed sub components
•
•

•

Job tracker: Centrally monitors the submitted Job and controls all processes running on the
nodes(computers) of the clusters.This communicated with Name Node for file system access
Task Tracker: Take care of the local job execution on the local file segments. This talks to DataNode for
file information. This constantly communicates with job Tracker daemon to report the task progress.

When the Hadoop system is running in a distributed mode all the daemons would
be running in the respective computer.

Hadoop Operational Modes
Hadoop can be run in one of the three modes
• Standalone(Local) Mode
– No daemons launched
– Everything runs in single JVM
– Suitable for development

•

Pseudo Distributed Mode
– All daemons are launched
– Runs on a single machine thus simulating a cluster environment
– Suitable for testing and debugging

•

Fully Distributed Mode
– The Hadoop daemons runs in a cluster environment
– Each daemons run on machine respectively assigned to them
– Suitable for integration Testing/Production

Hadoop Configuration Files
Filename
hadoop-env.sh

core-site.xml

hdfs-site.xml

mapred-site.xml
masters
slaves
hadoopmetrics.properties
log4j.properties

Format
Description
Bash script
Environment variables that are used in the scripts to run Hadoop
Hadoop
configuration.X Configuration settings for Hadoop Core, such as I/O settings that are
ML
common to HDFS and MapReduce
Hadoop
configuration.X Configuration settings for HDFS daemons: the namenode, the secondary
ML
namenode, and the datanodes
Hadoop
configuration.X Configuration settings for MapReduce daemons: the jobtracker, and the
ML
tasktrackers
Plain text
A list of machines (one per line) that each run a secondary namenode
Plain text
Java
Properties
Java
Properties

A list of machines (one per line) that each run a datanode and a tasktracker
Properties for controlling how metrics are published in Hadoop
Properties for system logfiles, the namenode audit log, and the task log for
the tasktracker child process

Key Configuration Properties
Property Name

Conf File

Standalone

Fs.default.name

core -site.xml

File:///(default) hdfs://localhost/

dfs.replication

hdfs -site.xml

NA

mapred.job.tracker mapes-site.xml local(default)

Pseudo Distributed Fully Distributed

hdfs://namenode/

1 3(default)

localhost:8021

jobtracket:8021

Design of HDFS
•
•
•
•
•

HDFS is a Hadoop's Distributed File System
Designed for storing very large files(Petabytes)
Single file can be stored across several disk
Not suitable for low latency data access
Designed to be highly fault tolerant hence can
run on commodity hardware

HDFS Concepts
• Like any file system HDFS stores files by
breaking it into smallest units called Blocks.
• The default HDFS block size is 64 MB
• The large block size helps in maintaining high
throughput
• Each block is replicated across multiple
machine in the cluster for redundancy.

Design of HDFS -Daemons
• The HDFS file system is managed by two daemons
• NameNode and DataNode
• NameNode and DataNode function in
master/slave fashion
– NameNode Manages File system namespace
– Maintains file system and metadata of all the files and
directories
• Namespace image
• Edit log

Design of HDFS –Daemons (cont.)
• Datanodes store and retrieve the blocks for the files when
they are told by NameNode.
• NameNode maintains the information on which DataNodes
all the blocks for a given files are located.
• DataNodes report to NameNode periodically with the list of
blocks they are storing.
• With NameNode off ,the HDFS is inaccessible.
• Secondary NameNode
– Not a backup for NameNode
– Just helps in merging namespace image with edit log to avoid
edit log becoming too large

Live Horizontal Scaling + Rebalancing

Configuring conf/*-site.xml files
• Need to set the hadoop.tmp.dir parameter
to a directory of your choice.
• We will use /app/hadoop/tmp
• sudo mkdir -p /app/hadoop/tmp
• sudo chmod 777 /app/hadoop/tmp

Configuring HDFS:

core-site.xml(Pseudo Distributed
Mode)

<?xml version ="1.0"?>

<configuration>
<property>
<name>hadoop.tmp.dir</name>
<value>/app/hadoop/tmp</value>
</property>
<property>
<name>fs.default.name</name>
<value>hdfs://localhost/</value>
</property>
</configuration>
Note: Add "fs.default.name" property under configuration tag to specify NameNode
location.
"localhost" for pseudo distributed mode. Name node runs at port 8020 by default if
no port specified.

Configuring mapred-site.xml(Pseudo Distributed Mode)
<?xml version="1.0"?>
<!— mapred-site.xml —>
<configuration>
<property>
<name>mapred.job.tracker</name>
<value>localhost:8021</value>
</property>
</configuration>
Note:Add "mapred.job.tracker" property under configuration tag to specify
JobTracker location.
"localhost:8021" for Pseudo distributed mode.
Lastly set JAVA_HOME in conf/hadoop-env.sh export
JAVA_HOME=/usr/lib/jvm/java-6-openjdk-amd64

Configuring HDFS: hdfs-site.xml(Pseudo Distributed
Mode)

• <?xml version ="1.0"?>
• 
<configuration>
<property>
<name>dfs.replication</name>
<value>1</value>
</property>
</configuration>

Starting HDFS
•

•

•

Format NameNode
hadoop namenode -format
– creates empty file system with storage directories and persistent data structure
– Data nodes are not involved
Start dfs service
– start-dfs.sh
– Verify daemons :
jps
If you get the namespace exception, copy the namespace id of the namenode
Paste it in the : /app/hadoop/tmp/dfs/data/current/VERSION file
– Stop : stop-dfs.sh
List/check HDFS
hadoop fsck / -files -blocks
hadoop fs -ls
hadoop fs -mkdir testdir
hadoop fs -ls

Verify HDFS
• Stop dfs services
stop-dfs.sh
Verify Daemons :jps
No java processes should be running

Configuration HDFS -hdfs-site
.xml(Pseudo Distributed Mode)
Property
Name
Description

Default Value

dfs.name.dir

Directories for NameNode to store it's
persistent data(Comma seperated directory
name).A copy of metadata is stored in each of
listed directory
${hadoop.tmp.dir}/dfs/name

dfs.data.dir

Directories where DataNode stores
blocks.Each block is stored in only one of
these directories

fs.checkpoint.dir

${hadoop.tmp.dir}/dfs/data

Directories where secondary namenode
stores checkpoints.A copy of the checkpoint is
stored in each of the listed directory
${hadoop.tmp.dir}/dfs/namesecondary

Basic HDFS Commands
•

Creating Directory
hadoop fs -mkdir <dirname>

•

Removing Directory
hadoop fs -rm <dirname>

•

Copying files to HDFS from local file system
hadoop fs -put <local dir>/<filename> <hdfs dir Name>/

•

Copying files from HDFS to local file system
hadoop fs -get <hdfs dir Name>/<hdfs file Name> <local dir>/

•

List files and directories
hadoop fs -ls <dir name>

•

List the blocks that make up each files in HDFS

HDFS Web UI
• Hadoop provides a web UI for viewing HDFS
– Available at http://namenode-host-ip:50070/
– Browse file system
– Log files

MapReduce
• A distributed parallel processing engine of Hadoop
• Processes the data in sequential parallel steps called
• Map
• Reduce
• Best run with a DFS supported by Hadoop to exploit it's parallel processing
abilities
• Has the ability to run on a cluster of computers
• Each computer called as a node
• Input and output data at every stage is handled in terms of key / value pairs
• Key / Value can be choose by programmer
• Mapper output is always sorted by key
• Mapper output with the same key are sent to the same reducer
• Number of mappers and reducers per node can be configured

Start the Map reduce daemons
• start-mapred.sh

The Overall MapReduce Word count Process
Input

Splitting

Mapping

Shuffling

Reducing

Final Result

Job Submission – Reduce Phase

Anatomy of a MapReduce program

mapred-site. xml - Pseudo Distributed Mode
<?xml version="1.0"?>
<!— mapred-site.xml —>
<configuration>
<property>
<name>mapred.job.tracker</name>
<value>localhost:8021</value>
</property>
</configuration>
Note:Add "mapred.job.tracker" property under configuration tag to specify JobTracker
location.
"localhost:8021" for Pseudo distributed mode.
Lastly set JAVA_HOME in conf/hadoop-env.sh
export JAVA_HOME=/usr/lib/jvm/java-6-openjdk-amd64

Starting Hadoop -MapReduce Daemons

• Start MapReduce Services
start-mapred.sh
– Verify Daemons:jps

MapReduce Programming
• Having seen the architecture of MapReduce, to
perform a job in hadoop a programmer need to create
• A MAP function
• A REDUCE function
• A Driver to communicate with the framework,
configure and launch the job

Map Function
•

The Map function is represented by Mapper class, which declares an abstract
method map()
• Mapper class is generic type with four type parameters for the input and output
key/ value pairs
•
Mapper <k1, v1, k2, v2>
•
kl, vi are the types of the input key / value pair
•
k2, v2 are the types of the output key/value pair
• Hadoop provides it's own types that are optimized for network serialization
• Text Corresponds to Java String
• LongWritable Corresponds to Java Long
• IntWritable Corresponds to Java Int
• The map() method must be implemented to achieve the input key / value
transformation
• map method is called by MapReduce framework passing the input key values from
the input file
• map method is provided with a context object to which the transformed key values
can be written to

Mapper — Word Count

•
•

public static class TokenizerMapper
extends Mapper<LongWritable, Text, Text, IntWritable> {

•
•
•
•
•
•
•
•
•
•
•
•
•
•

private final static IntWritable one = new IntWritable(1);
private Text word = new Text();
@Override
public void map(LongWritable key, Text value, Context context)
throws 1OException, InterruptedException {
StringTokenizer itr = new StringTokenizer(value,toString0, " tnrf,.;:?[]");
while (itr.hasMoreTokens()) {
word.set(itr.nextToken0.toLowerCase();
context.write(word, one);
}
}

Reduce Function
•
•

•
•
•
•

The Reduce function is represented by Reducer class, which declares an
abstract method reduce()
Reducer class is generic type with four type parameters for the input and output
key/ value pairs
• Reducer <k2, v2, k3, v3>
• k2, v2 are the types of the input key / value pair, this type of this pair must
match the output types of Mapper
• k3, v3 are the types of the output key / value pair
The reduce() method must be implemented to achieve the desired transformation
of input key / value
Reduce method is called by MapReduce framework passing the input key values
from out of map phase
MapReduce framework guarantees that the records with the same keys from all
the map tasks will reach a single reduce task
Similar to the map, reduce method is provided with a context object to which the
transformed key values can be written to

Reducer — Word Count

•
•
•
•
•
•
•
•
•
•
•
•
•

public static class IntSumReducer
extends Reducer< Text, IntWritable, Text, IntWritable> {
@Override
public void reduce(Text key, Iterable<IntWritable> values, Context context)
throws I0Exception, InterruptedException
{
int sum = 0;
for (IntWritable value values) {
sum += value.get();
}
context.write(key, new IntWritable(sum));
}
}

MapReduce Job Driver— Word Count
Public class WordCount (
public static void main(String args[])) throws Exception {
if (args.length 1=2) {
System.errprintln("Usage: WordCount <input Path> <output Path>");
System.exit(-1); }
Configuration conf- new Configuration();
Job job = new Job(conf, "Word Count");
job.setJarByClass(WordCount.class);
job.setMapperClass(TokenizerMapperclass);
job.setReducerClass(IntSumReducer.class);
job.setMapOutputKeyClass(Textclass);
job.setMapOutputValueClass(IntWritable.class);
job.setOutputKeyClass(Text.class);
job.setOutputValueClass(IntWritable.class);
FllelnputFormat.addlnputPath(job, new Path(args[0]);
FlleOutputFormatsetOutputPathCob, new Path(args[1]);
System.exit(job.waitForCompletion(true) ? 0: 1);
}
}

MapReduce Job — Word Count
• Copy wordcount.jar from the Hard drive folder to the
Ubuntu VM
hadoop jar <path_to_jar>/wordcount.jar
com.hadoop.WordCount <hdfs_input_dir>/pg5000.txt
<hdfs_output_dir>
• The <hdfs_output_dir> must not exist
• To view the output directly:
hadoop fs –cat ../../part-r-00000
• To copy the result to local:
hadoop fs -get <part-r-*****>

The MapReduce Web UI

• Hadoop provides a web UI for viewing job
information
• Available at http://jobtracker-host:50030/
• follow job's progress while it is running
• find job statistics
• View job logs
• Task Details

Combiner
• Combiner function helps to aggregate the map
output before passing on to reduce function
• Reduces intermediate data to be written to disk
• Reduces data to be transferred over network
• Combiner for a job is specified as
job.setCombinerClass(<combinerclassname>.clas
s);
• Combiner is represented by same interface as
Reducer

Word Count — With Combiner
Public class WordCount {
public static void main(String args[]) throws Exception{
if (args.length!=2) {
System. err printIn("Usage: WordCount <input Path> <output

Path>");
System.exit(-1);
}
Configuration conf = new Configuration();
Job job = new Job(conf, "Word Count");
job.setJarByClass(WordCount.class);
job.setMapperClass(TokenizerMapperclass);
job.setComblnerClass(IntSumReducer.class);
job.setReducerClass(IntSurriReducer class);
job.setMapOutputKeyClass(Text.class);
job.setMapOutputValueClass(IntWritable.class);
job.setOutputKeyClass(Text.class);
job.setOutputValueClass(IntWritable.class);
FilelnputFormataddlnputPath(job, new Path(args[0]) );
FileOutputFormat.setOutputPath(job, new Path(args[1]) );
System.exit(job.wait-ForCompletion(true) ? 0 : 1);

• hadoop jar lineindexer.jar
com.hadoop.LineIndexer <hdfs_input_dir>
<hdfs_output_dir>
• hadoop fs –cat ../../part-r-*****

• hadoop fs -get <part-r-*****>

Partitioning

• Map tasks partition their output keys by the number of reducers
• There can be many keys in a partition
• All records for a given key will be in a single partition
• A Partitioner class controls partitioning based on the Key
• Hadoop uses hash partition by default (HashPartitioner)
• The default behavior can be changed by implementing the getPartition()
method in the Partitioner (abstract) class
public abstract class Partitioner<KEY, VALUE> {
public abstract int getPartition(KEY key, VALUE value, int
numPartitions);
}
• A custom partitioner for a job can be set as
Job.setPartitionerClass(<customPartitionerClass>.class);

Partitioner— redirecting output from Mapper

Partitioner Example

public class WordPartitioner extends Partitioner <Text, IntWritable>{
@Override
public int getPartition(Text key, Int-Writable value,int numPartitions){
String ch = key.toStringasubstring(0,1);
/*if (ch.matches("[abcdefghijklm]"))
return 0;
}
else if (ch.matches("[Inopqrstuvwxyzn]")){
return 1;
}
return 2; */

//return (ch.charAt(0) % numPartitions); //round robin based on ASCI value return 0;
return 0;// default behavior
}
}

Reducer progress during Mapper
• MapReduce job shows something like Map(50%)
Reduce(10%)
• Reducers start copying intermediate key-value pairs
from the mappers as soon as they are available. The
progress calculation also takes in account the
processing of data transfer which is done by reduce
process, therefore the reduce progress starts showing
up as soon as any intermediate key-value pair for a
mapper is available to be transferred to reducer.
• The programmer defined reduce method is called only
after all the mappers have finished.

Hadoop Streaming
• Hadoop Streaming is a utility which allows
users to create and run jobs with any
executables (e.g. shell utilities) as the mapper
and/or the reducer.
• Using the streaming system you can develop
working hadoop jobs with extremely limited
knowldge of Java
• Hadoop actually opens a process and writes
and reads from stdin and stdout

Hadoop Streaming
• Hadoop Streaming uses Unix standard streams
as the interface between Hadoop and your
program, so you can use any language that
• Can read standard input and write to standard
output to write your MapReduce program.
• Streaming is naturally suited for text
processing.

Hands on
• Folder :
– examples/Hadoop_streaming_python
– Files “url1” & “url2’ are the input
– multifetch.py is the mapper (open it)
– reducer.py is the reducer(open this as well)

Hands on
• Run :
hadoop fs -mkdir urls
hadoop fs -put url1 urls/
hadoop fs -put url2 urls/
hadoop jar contrib/streaming/hadoop-streaming-1.0.3.jar
-mapper <path>/multifetch.py
-reducer <path>reducer.py
-input urls/*
-output titles

Decomposing problems into M/R jobs
• Small map-reduce jobs are usually better
– Easier to implement, test & maintain
– Easier to scale & reuse

• Problem : Find the word/letter that has the
maximum occurrences in a set of documents

Decomposing
• Count of each word/letter

M/R job (Job 1)

• Find max word/letter count M/R job (Job 2)

Choices can depend on complexity of jobs

Job Chaining
• Multiple jobs can be run in a linear or complex dependent fashion

Simple Dependency /Linear Chain

Directed Acyclic Graph(DAG)

• Simple way is to call the job drivers one after the other with respective
configurations
JobClient.runJob(conf1);
JobClient.runJob(conf2);
• If a job fails, the runJob() method will throw an IOException, so later jobs
in the pipeline don’t get executed.

Job Chaining
For complex dependencies you can use JobControI, and
ControlledJob classes
Controlledjob cjob1 = new ControlledJob(conf1);
ControlledJob cjob2 = new ControlledJob(conf2);
cjob2.addDependingjob(cjob1);
JobControl jc = new JobControl("Chained Job");
jc.addjob(cjob1);
jc.addjob(cjob2);
jc.run();

Apache Oozie
•
•
•
•
•
•
•

Work flow scheduler for Hadoop
Manages Hadoop Jobs
Integrated with many Hadoop apps i.e. Pig
Scalable
Schedule jobs
A work flow is a collection of actions i.e.
– map/reduce, pig
A work flow is
– Arranged as a DAG ( direct acyclic graph )
– Graph stored as hPDL ( XML process definition )

Oozie
• Engine to build complex DAG workflows
• Runs in it’s own daemon
• Describe workflows in set of XML &
configuration files
• Has coordinator engine that schedules
workflows based on time & incoming data
• Provides ability to re-run failed portions of the
workflow

Need for High-Level Languages
• Hadoop is great for large-data processing!
– But writing Java programs for everything is
verbose and slow
– Not everyone wants to (or can) write Java code

• Solution: develop higher-level data processing
languages
– Hive: HQL is like SQL
– Pig: Pig Latin is a bit like Perl

Hive and Pig
• Hive: data warehousing application in Hadoop
– Query language is HQL, variant of SQL
– Tables stored on HDFS as flat files
– Developed by Facebook, now open source

• Pig: large-scale data processing system
– Scripts are written in Pig Latin, a dataflow language
– Developed by Yahoo!, now open source
– Roughly 1/3 of all Yahoo! internal jobs

• Common idea:
– Provide higher-level language to facilitate large-data processing
– Higher-level language “compiles down” to Hadoop jobs

Hive: Background
• Started at Facebook
• Data was collected by nightly cron jobs into
Oracle DB
• “ETL” via hand-coded python
• Grew from 10s of GBs (2006) to 1 TB/day new
data (2007), now 10x that

Source: cc-licensed slide by Cloudera

Hive Components
•
•
•
•

Shell: allows interactive queries
Driver: session handles, fetch, execute
Compiler: parse, plan, optimize
Execution engine: DAG of stages (MR, HDFS,
metadata)
• Metastore: schema, location in HDFS, SerDe


Data Model
• Tables
– Typed columns (int, float, string, boolean)
– Also, list: map (for JSON-like data)

• Partitions
– For example, range-partition tables by date

• Buckets
– Hash partitions within ranges (useful for sampling,
join optimization)


Metastore
• Database: namespace containing a set of
tables
• Holds table definitions (column types, physical
layout)
• Holds partitioning information
• Can be stored in Derby, MySQL, and many
other relational databases


Physical Layout
• Warehouse directory in HDFS
– E.g., /user/hive/warehouse

• Tables stored in subdirectories of warehouse
– Partitions form subdirectories of tables

• Actual data stored in flat files
– Control char-delimited text, or SequenceFiles
– With custom SerDe, can use arbitrary format


Hive: Example
• Hive looks similar to an SQL database
• Relational join on two tables:
– Table of word counts from Shakespeare collection
– Table of word counts from the bible
SELECT s.word, s.freq, k.freq FROM shakespeare s
JOIN bible k ON (s.word = k.word) WHERE s.freq >= 1 AND k.freq >= 1
ORDER BY s.freq DESC LIMIT 10;
the
I
and
to
of
a
you
my
in
is

25848
23031
19671
18038
16700
14170
12702
11297
10797
8882

62394
8854
38985
13526
34654
8057
2720
4135
12445
6884

Hive: Behind the Scenes
SELECT s.word, s.freq, k.freq FROM shakespeare s
JOIN bible k ON (s.word = k.word) WHERE s.freq >= 1 AND k.freq >= 1
ORDER BY s.freq DESC LIMIT 10;

(Abstract Syntax Tree)
(TOK_QUERY (TOK_FROM (TOK_JOIN (TOK_TABREF shakespeare s) (TOK_TABREF bible k) (= (. (TOK_TABLE_OR_COL s) word) (.
(TOK_TABLE_OR_COL k) word)))) (TOK_INSERT (TOK_DESTINATION (TOK_DIR TOK_TMP_FILE)) (TOK_SELECT (TOK_SELEXPR (.
(TOK_TABLE_OR_COL s) word)) (TOK_SELEXPR (. (TOK_TABLE_OR_COL s) freq)) (TOK_SELEXPR (. (TOK_TABLE_OR_COL k) freq))) (TOK_WHERE
(AND (>= (. (TOK_TABLE_OR_COL s) freq) 1) (>= (. (TOK_TABLE_OR_COL k) freq) 1))) (TOK_ORDERBY (TOK_TABSORTCOLNAMEDESC (.
(TOK_TABLE_OR_COL s) freq))) (TOK_LIMIT 10)))

(one or more of MapReduce jobs)

Hive: Behind the Scenes

STAGE DEPENDENCIES:
Stage-1 is a root stage
Stage-2 depends on stages: Stage-1
Stage-0 is a root stage

STAGE PLANS:
Stage: Stage-1
Map Reduce
Alias -> Map Operator Tree:
s
TableScan
alias: s
Filter Operator
predicate:
expr: (freq >= 1)
type: boolean
Reduce Output Operator
key expressions:
expr: word
type: string
sort order: +
Map-reduce partition columns:
expr: word
type: string
tag: 0
value expressions:
expr: freq
type: int
expr: word
type: string
k
TableScan
alias: k
Filter Operator
predicate:
expr: (freq >= 1)
type: boolean
key expressions:
expr: word
type: string
sort order: +
Map-reduce partition columns:
expr: word
type: string
tag: 1
value expressions:
expr: freq
type: int

Stage: Stage-2
Map Reduce
Alias -> Map Operator Tree:
hdfs://localhost:8022/tmp/hive-training/364214370/10002
key expressions:
expr: _col1
type: int
sort order: tag: -1
value expressions:
expr: _col0
type: string
expr: _col1
type: int
expr: _col2
type: int
Reduce Operator Tree:
Extract
Limit
File Output Operator
compressed: false
GlobalTableId: 0
table:
input format: org.apache.hadoop.mapred.TextInputFormat
output format: org.apache.hadoop.hive.ql.io.HiveIgnoreKeyTextOutputFormat

Reduce Operator Tree:
Join Operator
condition map:
Inner Join 0 to 1
condition expressions:
0 {VALUE._col0} {VALUE._col1}
1 {VALUE._col0}
outputColumnNames: _col0, _col1, _col2
Filter Operator
predicate:
Stage: Stage-0
expr: ((_col0 >= 1) and (_col2 >= 1))
Fetch Operator
type: boolean
limit: 10
Select Operator
expressions:
expr: _col1
type: string
expr: _col0
type: int
expr: _col2
type: int
outputColumnNames: _col0, _col1, _col2
File Output Operator
compressed: false
GlobalTableId: 0
table:
input format: org.apache.hadoop.mapred.SequenceFileInputFormat
output format: org.apache.hadoop.hive.ql.io.HiveSequenceFileOutputFormat

Example Data Analysis Task
Find users who tend to visit “good” pages.
Visits

Pages

url

time

url

pagerank

Amy

www.cnn.com

8:00

www.cnn.com

0.9

Amy

www.crap.com

8:05

www.flickr.com

0.9

Amy

www.myblog.com

10:00

www.myblog.com

0.7

Amy

www.flickr.com

10:05

www.crap.com

0.2

Fred

cnn.com/index.htm 12:00

...

user

...

System-Level Dataflow
Visits

load

Pages

...

...

load

canonicalize

join by url

...
group by user

...
the answer
Pig Slides adapted from Olston et al.

compute average pagerank
filter

MapReduce Code
i
i
i
i

m
m
m
m

p
p
p
p

o
o
o
o

r
r
r
r

t
t
t
t

j
j
j
j

a
a
a
a

v
v
v
v

a
a
a
a

.
.
.
.

i
u
u
u

o
t
t
t

.
i
i
i

I
l
l
l

O
.
.
.

E
A
I
L

x
r
t
i

c
r
e
s

e
a
r
t

p t i o n ;
y L i s t ;
a t o r ;
;

r e p o r t e r . s e t S t a t u s ( " O K " ) ;
}
/ /
f o r

i m p o r t
o r g . a p a c h e . h a d o o p . f s . P a t h ;
i m p o r t
o r g . a p a c h e . h a d o o p . i o . L o n g W r i t a b l e ;
i m p o r t
o r g . a p a c h e . h a d o o p . i o . T e x t ;
i m p o r t
o r g . a p a c h e . h a d o o p . i o . W r i t a b l e ;
im p o r t
o r g . a p a c h e . h a d o o p . i o . W r i t a b l e C o m p a r a b l e ;
i m p o r t
o r g . a p a c h e . h a d o o p . m a p r e d . F i l e I n p u t F o r m a t
i m p o r t
o r g . a p a c h e . h a d o o p . m a p r e d . F i l e O u t p u t F o r m a
i m p o r t
o r g . a p a c h e . h a d o o p . m a p r e d . J o b C o n f ;
i m p o r t
o r g . a p a c h e . h a d o o p . m a p r e d . K e y V a l u e T e x t I n p
i m p o r t
o r g . a . h a d o o p . m a p r e d . M a p p e r ;
p a c h e
i m p o r t
o r g . a p a c h e . h a d o o p . m a p r e d . M a p R e d u c e B a s e ;
i m p o r t
o r g . a p a c h e . h a d o o p . m a p r e d . O u t p u t C o l l e c t o r
i m p o r t
o r g . a p a c h e . h a d o o p . m a p r e d . R e c o r d R e a d e r ;
i m p o r t
o r g . a p a c h e . h a d o o p . m a p r e d . R e d u c e r ;
i m p o r t
o r g . a p a c h e . h a d o o p . m a p r e d . R e p o r t e r ;
i mo r t
p
o r g . a p a c h e . h a d o o p . m a p r e d . S e q u e n c e F i l e I n p u
i m p o r t
o r g . a p a c h e . h a d o o p . m a p r e d . S e q u e n c e F i l e O u t
i m p o r t
o r g . a p a c h e . h a d o o p . m a p r e d . T e x t I n p u t F o r m a t
i m p o r t
o r g . a p a c h e . h a d o o p . m a p r e d . j o b c o n t r o l . J o b ;
i m p o r t
o r g . a p a c h e . h a d o o p . m a p roendt.rjoolb;c o n t r o l . J o b C
i m p o r t
o r g . a p a c h e . h a d o o p . m a p r e d . l i b . I d e n t i t y M a p

/
S
i
S
S
T
/
/
T
o

/
t
n
t
t
e
/
/
e
c

r
t
r
r
x
x
.

v o i
O u t
R e
P u l
i n g
f i
i n g
i n g
t
o
P r e
i t
t
o
c o l

d
m a
p u t C
p o r t
l
t h
l i n
r s t C
k e y
s t
v a l
u t K e
p e n d
c a m e
u t V a
l e c t

p
o
e
e
e
o
r
u
y

( L o n g W r i t a b
l l e c t o r < T e x
r
r e p o r t e r )
k e y
o u t
=
v a l . t o S t
m m a
=
l i n e .
= n l i n e . s u b s
i
g ( 0 ,
f i r
e
=
l i n e . s u
=
n e w
T e x t
a n
i n d e x
t o
f r o m .
l" =+ nveawl uTee)x;t
( o u t K e y ,
o u t

;

p e r ;

,

}
p
p t i o
r i t a

o m m a
w e

k

( " 1
V a l ) ;

}
}
p u b l i c
s t a t i c
c l a s s
L o a d A n d F i l t e r U s e r s
e x t e n d s
M
i m p l e m e n t s
M a p p e r < L o n g W r i t a b l e ,
T e x t ,
T e x t ,
l ,
e p t i o
}
p
T e x t >

t C o m m
w
;

}
p u b l i c
s t a t i c
c l a s s
J o i n
e x t e n d s
M a p R e d u c e B a s e
i m p l e m e n t s
R e d u c e r < T e x t ,
T e x t ,
T e x t ,
T e x t >
p u b l i c

/ /
s t o r e

v o
I
O
R
F o

i
t
u
e
r

d
r e
e r a t
t p u t
p o r t
e a c

d
o
C
e
h

u c e ( T
r < T e x
o l l e c
r
r e p
v a l u

e
t
t
o
e

x t
k e y
>
i t e r
o r < T e x
r t e r )
,
f i g u

}
p
{

,
,
t ,
T e x t >
o c ,
t h r o w s
I O E x c e p t i o
r e
o u t
w h i c h
f i l e

i t
/ /
a c c o r d i n g l y .
L i s t < S t r i n g >
f i r s t
=
n e w
A r r a y L i s t < S t r i n
L i s t < S t r i n g >
s e c o n d
=
n e w
A r r a y L i s t < S t r i

w h i l e
T
S
i
f i r s t . a d d ( v a l u e . s
e

e
t
f
u
l

( i t
x t
r i n
( v
b s t
s e

e r . h a s N e x t
t
=
i t e r . n
g tv a l u e )=
S
r i n g (
;
a l u e . c h a r A
r i n g ( 1 ) ) ;
s e c o n d . a d d

Pig Slides adapted from Olston et al.

(
e
t
t

)
x
.
(

)
{
t ( ) ;
t o
0 )
= =

s
:
g

p r o d u c t
f i r s t )
s 2
:
s e
o u t v a l
=
l e c t ( n u l l
e r . s e t S t a

l p . s e t O u t p u t K e y C l
l p . s e t O u t p u t V a l u e
l p . s e t M a p p e r C l a s s
t h e
v a l u e s e I n p u t F o r m a t . a
F i l
P a t h ( " / r / g a t e s / p a g e s " ) ) ;
u s e
d )
{
F i l e O u t p u t F o r m a t .
y
+
" , "
+
s 1
+
" , "
+
s 2 ;
n e w
P a t h ( " / u s
e w
T e x t ( o u t v a l ) ) ; l p . s e t N u m R e d u c e T a
( " O K " ) ;
J o b
l o a d P a g e s
=
n

a n d
{
c o
k
,
t u

n
e
n
s

c o l l e c t

}
p
' 1 ' )

( v a l u e . s u b s t r i n g ( 1 ) ) ;

J
e
l
l
e
l
e
l
l
F
P a t h ( " / u s
F

o
t
f
f
f
>
f
f
i
e
i

b
J
u
u
u
u
u
l
r
l

C
o
.
.
.
{
.
.
e
/
e

o
b
s
s
s
s
s
I
g
O
n
{ l f u . s
J o b
l

n
N
e
e
e
e
n
a
u
e
e
o

f
a
t
t
t
t
p
t
t
w
t
a

a
C
(
d

s
l
L
d

s
a
o
I

(
s
a
n

T
s
d
p

e
(
P
u

s
e
s
e

e t O u t
r / g a t
k s ( 0 )
w
J o b

l f u
=
n e w
J o b C o
e ( " L o a d
a n d
F i l
n p u t F o r m a t ( T e x t
u t p u t K e y C l a s s ( T
u t p u t V a l u e C l a s s
a p p e r C l a s s ( L o a d
t F o r m a t . a d d f u ,
I n p u t P a t h ( l
s / u s e r s " ) ) ;
u t F o r m a t . s e t O u t
P a t h ( " / u s e r / g a t
N u m R e d u c e T a s k s ( 0
d U s e r s
=
n e w
J o b

m
I
O
O
M
u
e
p

x
T
a
t

t
e
g
P

.
x
e
a

c
t
s
t

l
.
.
h

a
c
c
(

s
l
l
l

s
a
a
p

) ;
s s ) ;
s s ) ;
,
n e w

p u t P a t h ( l p ,
e s / t m p / i n d e x e
;
( l p ) ;
n f ( M
t e r
I n p u
e x t .
( T e x
A n d F
n e w

R
U
t
c
t
i

E
s
F
l
.
l

x
e
o
a
c
t

a
r
r
s
l
e

m
s
m
s
a
r

p
"
a
)
s
U

l
)
t
;
s
s

e .
;
. c

) ;
e r
v o i d
m a p (
T e x t
k ,
T e x t
v a l ,
p u t P a t h ( l f u ,
O u t p u t C o l l e L o n g W r i t a b l e >
c t o r < T e x t ,
o c ,
e s / t m p / f i l t e r
a t ;
R e p o r t e r
r e p o r t e r )
t h r o w s
I O E x c e p t i o n
) ;
r m a t ;
/ /
F i n d
t h e
u r l
( l f u ) ;
S t r i n g
l i n e
=
v a l . t o S t r i n g ( ) ;
i n t
f i r s t C o m m a
=
l i n e . i n d e x O f ( ' , ' ) ;
J o b C o n f
j o i n
= Rn e w mJ o b C o n f ( s ) ;
M
E x a
p l e . c l a s
i n t
s e c o n d C o m m a
=
l i n e . i n d e x O f ( ' , ' ,
C o m m a ) ;
f i r s t
j o i n . s e t J o b N a m e ( " J o i n
U s e r s
a n d
P a g e
S t r i n g
k e y
=
l i n e . s u b s t r i n g ( f i r s t C o m m a ,
s e c o n d C o m m a ) ; p u t F o r m a t ( K e y V a l u e T e x t I n p u
j o i n . s e t I n
/ /
d r o p
t h e
r e s t
o f
t h e
r e c o r d ,
I
d o n ' t
n e e d o i t . a n y m o r e , t K e y C l a s s ( T e x t . c l a s s ) ;
j
n
s e t O u t p u
/ /
j u s t
p a s s
a
1
f o r
t h e
c o m b i n e r / r e d u c e r
t o o s u m s i n s t e a d . V a l u e C l a s s ( T e x t . c l a s s )
j
i n .
e t O u t p u t
a s e
T e x t
o u t K e y
=
n e w
T e x t ( k e y ) ;
j o i n . s e t M a p p e r C lp e r . c l a s s ) ; y M a p
a s s ( I d e n t i t
T e x t >
{ c . c o l l e c t ( o u t K e y ,
o
n e w
L o n g W r i t a b l e ( 1 L ) ) ;
j o i n . s e t R e d u c e r C l a s s ( J o i n . c l a s s ) ;
}
F i l e I n p u t F o r m a t . a d d I n p u t P a t h ( j o i n ,
n
P a t h ( " / u s e r / g a t e s / t m p / i n d e x e d _ p a g e s " ) ) ;
u b l i c
s t a t i c
c l a s s
R e d u c e U r l s
e x t e n d s
M a p R e d u c e B a s e i l e I n p u t F o r m a t . a d d I n p u t P a t h ( j o i n ,
F
n
n
{ i m p l e m e n t s
R e d u c e r < T e x t ,
L o n g W r i t a b l e ,
W r i t a b l e C o m p a r a b l e , m p / f i l t e r e d _ u s e r s " ) ) ;
P a t h ( " / u s e r / g a t e s / t
b l e >
{
F i l e O u t p u t F o r m a t . s e ( j o i n ,
t O u t p u t P a t h
n e w
P a t h ( " / u s e r / g a t e s / t m p / j o i n e d " ) ) ;
p u b l i c
v o i d
r e d u c e (
j o i n . s e t N u m R e d u c e T a s k s ( 5 0 ) ;
y , T e x t
k e
J o b
j o i n J o b
=
n e w
J o b ( j o i n ) ;
+
1 ) ;
I t e r a t o r < L o n g W r i t a b l e >
i t e r ,
j o i n J o b . a d d D e p e n d i n g J o b ( l o a d P a g e s ) ;
O u t p u t C o l l e c t o r < W r i t a b l e C o m p a r a b l e ,
W r i t a b l e > b o c , d D e p e n d i n g J o b ( l o a d U s e r s ) ;
j o i n J o
. a d
n o w
w h i c h
f i l e o r t e r
R e p
r e p o r t e r )
t h r o w s
{
/ /
A d d
u p
a l l
t h e
v a l u e s
w e
s e e
J o b C o n f
g r o u p
= x a m p l e . c l a s s ) ; E
n e w
J o b C o n f ( M R
g r o u p . s e t J o b N a m e ( " G r o u p
U R L s " ) ;
l o n g
s u m
=
0 ;
g r o u p . s e t I n p u t F o r m a t ( K e y V a l u e T e x t I n p
i l e w(hi t e r . h a s N e x t ( ) )
{
g r o u p . s e t O u t p u t K e y C l a s s ( T e x t . c l a s s ) ;
s u m
+ =
i t e r . n e x t ( ) . g e t ( ) ;
g r o u p . s e t O u t p u t V a l u e C l a s s ( L o n g W r i t a b
a p R e d u c e B a s e e p o r t e r . s e t S t a t u s ( " O K " ) ;
r
g r o u p . s e t O u t p u t F o r m a t ( S e q u e n c e F i l a s s
l e O u t p u t F o r m a t . c
T e x t >
{ }
g r o u p . s e t M a p p e r C l a s s ( L o a d J o i n e d . c l a s
g r o u p . s e t C o m b i n e r C l a s s ( R e d u c e U r l s . c l
o c . c o l l e c t ( k e y ,
n e w
L o n g W r i t a b l e ( s u m ) ) ;
g r o u p . s e t R e d u c e r C l a s s ( R e d u c e U r l s . c l a
}
F i l e I n p u t F o r m a t . a d d I n p u t P a t h ( g r o u p ,
n
{
P a t h ( " / u s e r / g a t e s / t m p / j o i n e d " ) ) ;
u b l i c
s t a t i c
c l a s s
L o a d C l i c k s
e x t e n d s
M a p R e d u c e B a s e e O u t p u t F o r m a t . s e t O u t p u t P a t h ( g r o u p ,
F i l
m p lie m e n t s
M a p p e r < W r i t a b l e C o m p a r a b l e ,
W r i t aPbalteh,( "L/ounsgeWrr/igtaatbelse/,t m p / g r o u p e d " ) ) ;
{
g r o u p . s e t N u m R e d u c e T a s k s ( 5 0 ) ;
J o b
g r o u p J o b
=
n e w
J o b ( g r o u p ) ;
p u b l i c
v o i d
m a p (
g r o u p J o b . a d d D e p e n d i n g J o b ( j o i n J o b ) ;
W r i t a b l e C o m p a r a b l e
k e y ,
a ) ;
W r i t a b l e
v a l ,
J o b C o n f
t o p 1 0 0
=
n e w
J o b C o n f ( M R E x a m p
O u t p u t C o l l e c t o r < L o n g W r i t a b l e ,
T e x t >
o c , o p 1 0 0 . s e t J o b N a m e ( " T o p
t
1 0 0
s i t e s " ) ;
R e p o r t e r or e p o r t e r ) p t i o n
t h r
w s
I O E x c e
{
t o p 1 0 0 . s e t I n p u t F o r m a t ( S e q u e n c e F i l e I n
o c . c o l l e c t ( ( L o n g W r i t a b l e ) v a l ,
( T e x t ) k e y ) ;
t o p 1 0 0 . s e t O u t p u t K e y C l a s s ( L o n g W r i t a b l
}
t o p 1 0 0 . s e t O u t p u t V a l u e C l a s s ( T e x t . c l a s
t o p 1 0 0 . s e t O u t p u t F o r m a t ( S e q u e n c e F i l e O
o r m a t . c l a s s ) ;
u b l i c
s t a t i c
c l a s s
L i m i t C l i c k s
e x t e n d s
M a p R e d u c e B a s e p 1 0 0 . s e t M a p p e r C l a s s ( L o a d C l i c k s . c l a
t o
i m p l e m e n t s
R e d u c e r < L o n g W r i t a b l e ,
T e x t ,
L o n g W r i t a b l e , 0 T e x t > o { b i n e r C l a s s ( L i m i t C l i c k s .
t o p 1 0
. s e
C
m
t o p 1 0 0 . s e t R e d u c e r C l a s s ( L i m i t C l i c k s . c
i n t
c o u n t
=
0 ;
F i l e I n p u t F o r m a t . a d d I n p u t P a t h ( t o p 1 0 0 ,
p uvboliidc r e d u c e (
P a t h ( " / u s e r / g a t e s / t m p / g r o u p e d " ) ) ;
L o n g W r i t a b l e
k e y ,
F i l e O u t p u t F o r m a t . s e t O u t p u t P a t h ( t o p 1 0 0 ,
I t e r a t o r < T e x t >
i t e r ,
P a t h ( " / u s e r / g a t e s / t o p 1 0 0 s i t e s f o r u s e r s 1 8 t o 2 5 "
O u t p u t C o l l e c t o r < L o n g W r i t a b l e ,
T e x t >
o c ,
t o p 1 0 0 . s e t N u m R e d u c e T a s k s ( 1 ) ;
n
{
R e p o r t e r
r e p o r t e r )
t h r o w s
{
J o b
l i m i t
=
n e w
J o b ( t o p 1 0 0 ) ;
i t ' s
f r o m
a n d
l i m i t . a d d D e p e n d i n g J o b ( g r o u p J o b ) ;
/ /
O n l y
o u t p u t
t h e
f i r s t
1 0 0
r e c o r d s
w h i l e 1( c o u n t i t e r . h a s N e x t ( ) )
<
0 0
& &
{
J o b C o n t r o l
j c
=
n e w
J o b C o n t r o l ( " F i n d
1 0 0
s i t e s
f o r
g > ( ) ;
o c . c o l l e c t ( k e y ,
i t e r . n e x t ( ) ) ;
1 8
t o
2 5 " ) ;
n g > ( ) ;
c o u n t + + ;
j c . a d d J o b ( l o a d P a g e s ) ;
}
j c . a d d J o b ( l o a d U s e r s ) ;
}
j c . a d d J o b ( j o i n J o b ) ;
j c . a d d J o b ( g r o u p J o b ) ;
u b l i c
s t a t i c
v o i d
m a i n ( S t r i n g [ ]
a r g s )
t h r o w s
I O E x c e p t i o n J { b ( l i m i t ) ;
j c . a d d
o
J o b C o n f
l p
=
n e w
J o b C o n f ( M R E x a m p l e . c l a s s ) ;
j c . r u n ( ) ;
t J o b N a m e ( " L o a d
l p . s e
P a g e s " ) ;
}
l p . s e t I n p u t F o r m a t ( T e x t I n p u t F o r m a t . c l a s s ) ;
}
p u b l i c

t F o r m
p u t F o
;

l e
k ,
T e x t
v a l
t ,
T e x t >
o c ,
t h r o w s
I O E x c e
W
r i n g ( ) ;
i n d e x O f ( ' , ' ) ;
t C o m m a ) ;
b s t r i n g ( f i r s t C
( k e y ) ;
t h e
v a l u e
s o

p u b l i c
v o i d
m a p ( L o n g W r i t a b l e
k ,
T e x t
v a
O u t p u t C o l l e c t o r < T e x t ,
T e x t >
o c ,
R e p o r t e r
r e p o r t e r )
t h r o w s
I O E x c
/ /
P u l l
t h e
k e y
o u t
S t r i n g
l i n e
=
v a l . t o S t r i n g ( ) ;
i n t
f i r s t C o m m a
=
l i n e . i n d e x O f ( ' , ' ) ;
S t r i n g
v a l u e
=f ilrisnteC.osmumbas t+r i1n)g;(
i n t
a g e
=
I n t e g e r . p a r s e I n t ( v a l u e ) ;
i f
( a g e
<
1 8
| |
a g e
>
2 5 )
r e t u r n ;
S t r i n g
k e y
=
l i n e . s u b s t r i n g ( 0 ,
f i r s
T e x t
o u t K e y
=
n e w
T e x t ( k e y ) ;
/ /
P r e p e n d
a n
i n d ee
x
k n o w hw h i c h uf i l e
t o
t
e
v a l
e
s o
/ /
i t
c a m e
f r o m .
T e x t
o u t V a l
=
n e w
T e x t ( " 2 "
+
v a l u e )
o c . c o l l e c t ( o u t K e y ,
o u t V a l ) ;
}

r o s
s 1
r i n
i n g
c o l
o r t

}
}
}
}
u t F o r m a t ; c
p u b l i
s t a t i c
c l a s s
L o a d J o i n e d
e x t e n d s
M a p R e d u c e B a s
i m p l e m e n t s
M a p p e r < T e x t ,
T e x t ,
T e x t ,
L o n g W r i t a b l
;
t ;

p u b l i c
c l a s s
M R E x a m p l e
{
p u b l i c
s t a t i c
c l a s s
L o a d P a g e s
e x t e n d s
M a p R e d u c e B
i m p l e m e n t s
M a p p e r < L o n g W r i t a b l e ,
T e x t ,
T e x t ,
p u b l i c

D o
t h e
c
( S t r i n g
f o r
( S t
S t r
o c .
r e p

Pig Latin Script
Visits = load ‘/data/visits’ as (user, url, time);
Visits = foreach Visits generate user, Canonicalize(url), time;
Pages = load

‘/data/pages’ as (url, pagerank);

VP = join Visits by url, Pages by url;
UserVisits = group VP by user;
UserPageranks = foreach UserVisits generate user, AVG(VP.pagerank) as avgpr;
GoodUsers = filter UserPageranks by avgpr > ‘0.5’;
store GoodUsers into '/data/good_users';

HIVE
•
•
•
•

A datawarehousing framework built on top of Hadoop
Started at Facebook in 2006
Targets users are data analysts comfortable with SQL
Allows to query the data using a SQL-like language called
HiveQL
• Queries compiled into MR jobs that are executed on
Hadoop
• Meant for structured data

Hive Architecture cont.
• Can interact with Hive using :
•
•
•

CLI(Command Line Interface)
JDBC
Web GUI

• Metastore – Stores the system catalog and metadata about
tables, columns, partitions etc.
• Driver – Manages the lifecycle of a HiveQL statement as it
moves through Hive.
Query Compiler – Compiles HiveQL into a directed acyclic
graph of map/reduce tasks.
• Execution Engine – Executes the tasks produced by the
compiler interacting with the underlying Hadoop instance.
• HiveServer – Provides a thrift interface and a JDBC/ODBC
server.

Hive Architecture cont.
• Physical Layout
• Warehouse directory in HDFS
• e.g., /user/hive/warehouse

• Table row data stored in subdirectories of warehouse
• Partitions form subdirectories of table directories

• Actual data stored in flat files
• Control char-delimited text, or SequenceFiles

Hive Vs RDBMS
• Latency for Hive queries is generally very high (minutes) even
when data sets involved are very small
• On RDBMSs, the analyses proceed much more iteratively with
the response times between iterations being less than a few
minutes.

• Hive aims to provide acceptable (but not optimal) latency for
interactive data browsing, queries over small data sets or test
queries.
• Hive is not designed for online transaction processing and
does not offer real-time queries and row level updates.
It is best used for batch jobs over large sets of immutable data
(like web logs).

Supported Data Types
• Integers
•
•

BIGINT(8 bytes), INT(4 bytes), SMALLINT(2 bytes), TINYINT(1
byte).
All integer types are signed.

• Floating point numbers
•

FLOAT(single precision), DOUBLE(double precision)

• STRING : sequence of characters
• BOOLEAN : True/False

• Hive also natively supports the following complex types:
•
•
•

Associative arrays – map<key-type, value-type>
Lists – list<element-type>
Structs – struct<field-name: field-type, ... >

Hive : Install & Configure
• Download a HIVE release compatible with your Hadoop
installation from :
• http://hive.apache.org/releases.html
• Untar into a directory. This is the HIVE’s home directory
• tar xvzf hive.x.y.z.tar.gz
• Configure
• Environment variables – add in .bash_profile
• export HIVE_INSTALL=/<parent_dir_path>/hive-x.y.z
• export PATH=$PATH:$HIVE_INSTALL/bin
• Verify Installation
• Type : hive –help (Displays commands usage)
• Type : hive (Enter the hive shell)
hive>

Hive : Install & Configure cont.
•
•
•
•
•
•
•
•
•

Start Hadoop daemons (Hadoop needs to be running)
Configure to Hadoop
Create hive-site.xml in $HIVE_INSTALL/conf directory
Specify the filesystem and jobtracker using the properties
fs.default.name & mapred.job.tracker
If not set, these default to the local filesystem and the local(inprocess) job-runner
Create following directories under HDFS
/tmp (execute: hadoop fs –mkdir /tmp)
user/hive/warehouse (execute: hadoop fs –mkdir
/user/hive/warehouse)
chmod g+w for both (execute : hadoop fs –chmod g+w <dir_path>)

Hive : Install & Configure cont.
• Data Store
• Hive stores data under /user/hive/warehouse by default
• Metastore
• Hive by default comes with a light wight SQL database
Derby to store the metastore metadata.
• But this can be configured to other databases like MySQL
as well.
• Logging
• Hive uses Log4j
• You can find Hive’s error log on the local file system
at /tmp/$USER/hive.log

Hive Data Models
• Databases
• Tables
• Partitions
• Each Table can have one or more partition Keys which
determines how the data is stored.
• Partitions - apart from being storage units - also allow the user
to efficiently identify the rows that satisfy a certain criteria.
• For example, a date_partition of type STRING and
country_partition of type STRING.
• Each unique value of the partition keys defines a partition of the
Table. For example all “India" data from "2013-05-21" is a
partition of the page_views table.
• Therefore, if you run analysis on only the “India" data for 201305-21, you can run that query only on the relevant partition of
the table thereby speeding up the analysis significantly.

Partition example
CREATE TABLE logs (ts BIGINT, line STRING)
PARTITIONED BY (dt STRING, country STRING);
• When we load data into a partitioned table, the partition
values are specified explicitly:
LOAD DATA LOCAL INPATH 'input/hive/partitions/file1'
INTO TABLE logs
PARTITION (dt='2001-01-01', country='GB');

Hive Data Model cont.
• Buckets
• Data in each partition may in turn be divided into
Buckets based on the value of a hash function of some
column of the Table.
• For example the page_views table may be bucketed
by userid, which is one of the columns, other than the
partitions columns, of the page_views table. These
can be used to efficiently sample the data.

A Practical session
Starting the Hive CLI
• Start a terminal and run :
• $hive
• Will take you to the hive shell/prompt
hive>
• Set a Hive or Hadoop conf prop:
• hive> set propkey=propvalue;
• List all properties and their values:
• hive> set –v;

A Practical Session
Hive CLI Commands

• List tables:
– hive> show tables;
• Describe a table:
– hive> describe <tablename>;
• More information:
– hive> describe extended <tablename>;

A Practical Session
Hive CLI Commands
• Create tables:
– hive> CREATE TABLE cite (citing INT, cited INT)
>ROW FORMAT DELIMITED
>FIELDS TERMINATED BY ',' STORED AS TEXTFILE;

• The 2nd and the 3rd lines tell Hive how the data is stored (as a text
file)and how it should be parsed (fields are separated by commas).
• Loading data into tables
• Let’s load the patent data into table cite
hive> LOAD DATA LOCAL INPATH ‘<path_to_file>/cite75_99.txt’
> OVERWRITE INTO TABLE cite;

• Browse data
hive> SELECT * FROM cite LIMIT 10;

A Practical Session
Hive CLI Commands
•

Count

hive>SELECT COUNT(*) FROM cite;

Some more playing around
• Create table to store citation frequency of each patent
hive> CREATE TABLE cite_count (cited INT, count INT);

•

Execute the query on the previous table and store the results :

•
•
•

hive> INSERT OVERWRITE TABLE cite_count
> SELECT cited, COUNT(citing)
> FROM cite
> GROUP BY cited;

•

Query the count table

hive> SELECT * FROM cite_count WHERE count > 10 LIMIT 10;

•

Drop Table

hive> DROP TABLE cite_count;

Data Model
Partitioning Data
• One or more partition columns may be specified:
hive>CREATE TABLE tbl1 (id INT, msg STRING)
PARTITIONED BY (dt STRING);
• Creates a subdirectory for each value of the partition column, e.g.:
/user/hive/warehouse/tbl1/dt=2009-03-20/
• Queries with partition columns in WHERE clause will scan through
only a subset of the data.

Managing Hive Tables
• Managed table
• Default table created (without EXTERNAL keyword)
• Hive manages the data
CREATE TABLE managed_table (dummy STRING);
LOAD DATA INPATH '/user/tom/data.txt' INTO table managed_table;
• Moves the data into the warehouse directory for the table.

DROP TABLE managed_table;
• Deletes table data & metadata.

Managing Hive Tables
•

External table
• You control the creation and deletion of the data.
• The location of the external data is specified at table creation time:
• CREATE EXTERNAL TABLE external_table (dummy STRING)
• LOCATION '/user/tom/external_table';
LOAD DATA INPATH '/user/tom/data.txt' INTO TABLE external_table;
• DROP TABLE external_table
Hive will leave the data untouched and only delete the metadata.

Conclusion
• Supports rapid iteration of adhoc queries
• High-level Interface (HiveQL) to low-level infrastructure
(Hadoop).
• Scales to handle much more data than many similar
systems

Hive Resources/References
Documentation
• cwiki.apache.org/Hive/home.html
Mailing Lists
• hive-user@hadoop.apache.org
Books
• Hadoop, The Definitive Guide, 3rd edition by Tom White,
(O’Reilly)

PIG
• PIG is an abstraction layer on top of MapReduce that frees analysts from the
complexity of MapReduce programming
• Architected towards handling unstructured and semi structured data
• It's a dataflow language, which means the data is processed in a sequence of steps
transforming the data
• The transformations support relational-style operations such as filter, union, group,
and join.
• Designed to be extensible and reusable
• Programmers can develop own functions and use (UDFs)
• Programmer friendly
• Allows to introspect data structures
• Can do sample run on a representative subset of your input
• PIG internally converts each transformation into a MapReduce job and submits to
hadoop cluster
• 40 percent of Yahoo's Hadoop jobs are run with PIG

Pig : What for ?
• An ad-hoc way of creating and executing
map-reduce jobs on very large data sets
• Rapid development
• No Java is required
• Developed by Yahoo

Pig Use cases
 Processing of Web Logs.
 Data processing for search platforms.

 Support for Ad Hoc queries across large
datasets.
 Quick Prototyping of algorithms for
processing large datasets.

Use Case in Healthcare
Problem Statement:
De-identify personal health information.
Challenges:
 Huge amount of data flows into the systems daily and there are multiple data
sources that we need to aggregate data from.
 Crunching this huge data and deidentifying it in a traditional way had problems.

When to Not Use PIG
• Really nasty data formats or completely
unstructured data
(video, audio, raw human-readable text).

• Pig is definitely slow compared to Map Reduce
jobs.
• When you would like more power to optimize
your code.

PIG Architecture
• Pig runs as a client side application ,there is no
need to install anything on the cluster.

Install and Configure PIG
• Download a version of PIG compatible with your hadoop installation
• http://pig.apache.org/releases.html
• Untar into a designated folder. This will be Pig's home directory
• tar xzf pig-x.y.z.tar.gz
• Configure
• Environment Variables add in .bash_profile
• export PIG_INSTALL=/<parent directory path>/pig-x.y.z
• export PATH=$PATH:$PIG_INSTALL/bin
• Verify Installation
• Try pig -help
• Displays command usage
• Try pig
• Takes you into Grunt shell grunt>

PIG Execution Modes
• Local Mode
• Runs in a single JVM
• Operates on local file system
• Suitable for small datasets and for development
• To run PIG in local mode
• pig -x local
• MapReduce Mode
• In this mode the queries are translated into MapReduce jobs and
run on hadoop cluster
• PIG version must be compatible with hadoop version
• Set HADOOP HOME environment variable to indicate pig which
hadoop client to use
• export HADOOP_HOME=$HADOOP_INSTALL
• If not set it will use a bundled version of hadoop

Pig Latin Relational Operators

Ways of Executing PIG programs
Grunt
options
• Script

• An interactive shell for running Pig commands
• Grunt is started when the pig command is run without any

• Pig commands can be executed directly from a script file
pig pigscript.pig
• It is also possible to run Pig scripts from Grunt shell using run
and exec.
• Embedded
• You can run Pig programs from Java using the PigServer class,
much like you can use JDBC
• For programmatic access to Grunt, use PigRunner

An Example
Create a file sample.txt with(tab delimited) :

1932
1905

23
12

2
1

And so on.
grunt> records = LOAD ‘<your_input_dir>/sample.txt'
AS (year:chararray, temperature:int, quality:int);
DUMP records;
grunt>filtered_records = FILTER records BY temperature !=9999 AND
(quality ==0 OR quality ==1 OR quality ==4);

Example cont.
grunt>grouped_records = GROUP
filtered_records BY year;
grunt>max_temp = FOREACH grouped_records
GENERATE group,
MAX (filtered_records.temperature);
grunt>DUMP max_temp;

Data Types
• Simple Type
Category

Type

Description

int

32 - bit signed integer

long

64 - bit signed integer

float

32-bit floating-point number

double

64-bit floating-point number

Text

chararray

Character array in UTF-16 format

Binary

bytearray

Byte array

Numeric

Data Types
• Complex Types
Type

Description

Example

Tuple

Sequence of fields of any type

(1 ,'pomegranate' )

Bag

An unordered collection of tuples, possibly with
{(1 ,'pomegranate' ) ,(2)}
duplicates

map

A set of key-value pairs; keys must be character
arrays but values may be any type

['a'#'pomegranate' ]

LOAD Operator
<relation name> = LOAD '<input file with path>' [USING UDF()]
[AS (<field namel>:dataType, <field nome2>:dataType„<field
name3>:dataType)]
• Loads data from a file into a relation
• Uses the PigStorage load function as default unless specified otherwise with
the USING option
• The data can be given a schema using the AS option.
• The default data type is bytearray if not specified

records=LOAD 'sales.txt';
records=LOAD 'sales.txt' AS (fl:chararray, f2:int, f3:f1oat.);
records=LOAD isales.txt' USING PigStorage('t');
records= LOAD (sales.txt' USING PigStorage('t') AS (fl:chararray,
f2:int,f3:float);

Diagnostic Operators
• DESCRIBE
• Describes the schema of a relation
• EXPLAIN
• Display the execution plan used to compute
a relation
• ILLUSTRATE
• Illustrate step-by-step how data is
transformed
• Uses sample of the input data to simulate
the execution.

Data Write Operators
• LIMIT
• Limits the number of tuples from a relation
• DUMP
• Display the tuples from a relation
• STORE
• Store the data from a relation into a
directory.
• The directory must not exist

Relational Operators
• FILTER
• Selects tuples based on Boolean expression
teenagers = FILTER cust BY age <20;
• ORDER
• Sort a relation based on one or more fields
• Further processing (FILTER, DISTINCT, etc.) may
destroy the ordering
ordered list = ORDER cust BY name DESC;
• DISTINCT
• Removes duplicate tuples
unique_custlist = DISTINCT cust;

• GROUP BY
• Within a relation, group tuples with the same group key
• GROUP ALL will group all tuples into one group
groupByProfession=GROUP cust BY profession
groupEverything=GROUP cust ALL
• FOR EACH
• Loop through each tuple in nested_alias and generate new
tuple(s)
. countByProfession=FOREACH groupByProfession GENERATE
group, count(cust);
• Built in aggregate functions AVG, COUNT, MAX, MIN, SUM

• GROUP BY
• Within a relation, group tuples with the same group key
• GROUP ALL will group all tuples into one group
groupByProfession=GROUP cust BY profession
groupEverything=GROUP cust ALL
• FOR EACH
• Loop through each tuple in nested alias and generate new
tuple(s).
• At least one of the fields of nested alias should be a bag
• DISTINCT, FILTER, LIMIT, ORDER, and SAMPLE are allowed
operations in nested op to operate on the inner bag(s).
• countByProfession=FOREACH groupByProfession GENERATE
group, count(cust);
• Built in aggregate functions AVG, COUNT, MAX, MIN, SUM

Operating on Multiple datasets
• Join
Compute inner join of two or more relations based on common fields values
DUMP A;
(1,2,3)
(4,2,1)
(8,3,4)
(4,3,3)
(7,2,5)

DUMP B;
(2,4)
(8,9)
(1,3)
(2,7)
(7,9)

X=JOIN A BY a1,B BY b1
DUMP X;
(1,2,3,1,3)
(8,3,4,8,9)
(7,2,5,7,9)

• COGROUP
DUMP A;
(1,2,3)
(4,2,1)
(8,3,4)
(4,3,3)
(7,2,5)

Group tuples from two or more relations,based on common group values
DUMP B;
(2,4)
(8,9)
(1,3)
(2,7)
(7,9)

X=COGROUP A BY a1,B BY b1
DUMP X;
(1,{(1,2,3)},{(1,3)})
(8,{(8,3,4)},{(8,9)})
(7,{(7,2,5)},{(7,9)})
(2,{},{(2,4),(2,7)})
(4,{(4,2,1),(4,3,3)},{})

Joins & Cogroups
• JOIN and COGROUP operators perform similar functions.
• JOIN creates a flat set of output records while COGROUP creates
a nested set of output records

Data
File – student

File – studentRoll

Name

Age

GPA

Name

RollNo

Joe

18

2.5

Joe

45

3.0

Sam

24

Sam
Angel

21

7.9

Angel

1

John

17

9.0

John

12

Joe

19

2.9

Joe

19

Pig Latin - GROUP Operator
Example of GROUP Operator:
A = load 'student' as (name:chararray, age:int, gpa:float);
dump A;
(joe,18,2.5)
(sam,,3.0)
(angel,21,7.9)
(john,17,9.0)
(joe,19,2.9)

X = group A by name;
dump X;

(joe,{(joe,18,2.5),(joe,19,2.9)})
(sam,{(sam,,3.0)})
(john,{(john,17,9.0)})
(angel,{(angel,21,7.9)})

Pig Latin – COGROUP Operator
Example of COGROUP Operator:

A = load 'student' as (name:chararray, age:int,gpa:float);
B = load 'studentRoll' as (name:chararray, rollno:int);
X = cogroup A by name, B by name;
dump X;
(joe,{(joe,18,2.5),(joe,19,2.9)},{(joe,45),(joe,19)})
(sam,{(sam,,3.0)},{(sam,24)})
(john,{(john,17,9.0)},{(john,12)})
(angel,{(angel,21,7.9)},{(angel,1)})

UNION
Creates the union of two or more relations
DUMP A;
(1,2,3)
(4,2,1)
(8,3,4)

X=UNION A , B;
DUMP X;

(1,2,3)
(4,2,1)
(8,3,4)
(2,4)
(8,9)

DUMP B;
(2,4)
(8,9)

SPLITS

Splits a relation into two or more relations, based on a Boolean
expression

Y=SPLIT X INTO C IF a1<5 , D IF a1>5;
DUMP C;
DUMP D;
(1,2,3)
(4,2,1)
(8,3,4)
(2,4)
(8,9)

User Defined Functions (UDFs)
• PIG lets users define their own functions and lets them be
used in the statements
• The UDFs can be developed in Java, Python or Javascript
– Filter UDF
• To be subclassof FilterFunc which is a subclass ofEvalFunc

– Eval UDF
To be subclassed of EvalFunc
public abstract class EvalFunc<T> {
public abstract T exec(Tuple input) throws 1OException;
}

– Load / Store UDF
• To be subclassed of LoadFunc /StoreFunc

Creating UDF : Eval function example
public class UPPER extends EvalFunc<String> {
@Override

public String exec(Tuple input) throws IOException {
if (input == null || input.size() == 0)
return null;
try {
String str = (String) input.get(0);
return str.toUpperCase(); // java upper case
function
} catch (Exception e) {
throw new IOException("Caught exception processing
input...", e);
}
}
}

Define and use an UDF
• Package the UDF class into a jar
• Define and use a UDF
• REGISTER yourUDFjar_name.jar;
•
cust = LOAD some_cust_data;
• filtered = FOREACH cust GENERATE
•

com.pkg.UPPER(title) ;
DUMP filtered;

Piggy Bank
• Piggy Bank is a place for Pig users to share the
Java UDFs they have written for use with Pig.
The functions are contributed "as-is.“
• Piggy Bank currently supports Java UDFs.
• No binary version to download, download
source, build and use.

Pig Vs. Hive
• Hive was invented at Facebook.
• Pig was invented at Yahoo.
• If you know SQL, then Hive will be very
familiar to you. Since Hive uses SQL, you will
feel at home with all the
familiar select, where, group by, and order
by clauses similar to SQL for relational
databases.

Pig Vs. Hive
• Pig needs some mental adjustment for SQL
users to learn.
• Pig Latin has many of the usual data
processing concepts that SQL has, such as
filtering, selecting, grouping, and ordering, but
the syntax is a little different from SQL
(particularly the group
by and flatten statements!).

Pig Vs. Hive
• Pig gives you more control and optimization over
the flow of the data than Hive does.
• If you are a data engineer, then you’ll likely feel
like you’ll have better control over the dataflow
(ETL) processes when you use Pig Latin, if you
come from a procedural language background.
• If you are a data analyst, however, you will likely
find that you can work on Hadoop faster by using
Hive, if your previous experience was more with
SQL.

Pig Vs. Hive
• Pig Latin allows users to store data at
any point in the pipeline without disrupting
the pipeline execution.
• Pig is not meant to be an ad-hoc query tool.

What is Apache Mahout
• Mahout is an open source machine learning
library from Apache.
• Scalable, can run on Hadoop
• Written in Java
• Started as a Lucene sub-project, became an
Apache top-level-project in 2010.

250

Machine Learning : Definition
• “Machine Learning is programming computers to
optimize a performance criterion using example data or
past experience”
– Intro. To Machine Learning by E. Alpaydin
• Subset of Artificial Intelligence
• Lots of related fields:
– Information Retrieval
– Stats
– Biology
– Linear algebra
– Many more

Machine Learning
• Machine learning, a branch of artificial intelligence, concerns the
construction and study of systems that can learn from data.
• Supervised learning is tasked with learning a function from labeled
training data in order to predict the value of any valid input.
– Common examples of supervised learning include classifying e-mail messages as
spam.

• Unsupervised learning is tasked with making sense of data without any
examples of what is correct or incorrect. It is most commonly used for
clustering similar input into logical groups.

252

Common Use Cases
•
•
•
•
•

Recommend friends/dates/products
Classify content into predefined groups
Find similar content
Find associations/patterns in actions/behaviors
Identify key topics/summarize text
– Documents and Corpora

• Detect anomalies/fraud
• Ranking search results
• Others?

Apache Mahout
• An Apache Software Foundation project to create
scalable machine learning libraries under the Apache
Software License
– http://mahout.apache.org

• Why Mahout?

– Many Open Source ML libraries either:
•
•
•
•
•

Lack Community
Lack Documentation and Examples
Lack Scalability
Lack the Apache License
Or are research-oriented

Apache Mahout components

• The Three C’s:
– Collaborative Filtering (recommenders)
– Clustering
– Classification
• Others:
– Frequent Item Mining
– Math stuff

Content Discovery &
Personalization

Key points
• Automatic discovery of the most relevant content
without searching for it
• Automatic discovery and recommendation of the
most appropriate connection between people and
interests
• Personalization and presentation of the most
relevant content (content, inspirations,
marketplace, ads) at every page/touch point

Business objectives
• More revenue from incoming traffic
• Improved consumer interaction and loyalty, as
each page has more interesting and relevant
content

• Drives better utilization of assets within the
platform by linking “similar” products

Basic approaches to Recommendations
on social sites
• Collaborative Filtering (CF): Collaborative filtering first
computes similarity between two users based on their
preference towards items, and recommends items
which are highly rated(preferred) by similar users.
• Content based Recommendation(CBR) : Content
based system provides recommendation directly based
on similarity of items and the user history . Similarity is
computed based on item attributes using appropriate
distance measures.

Collaborative Filtering (CF)
– Provide recommendations solely based on preferences
expressed between users and items
– “People who watched this also watched that”

• Content-based Recommendations (CBR)
– Provide recommendations based on the attributes of the
items (and user profile)
– ‘Chak de India’ is a sports movie, Mohan likes sports
movies
• => Suggest Chak de India to Mohan

• Mahout geared towards CF, can be extended to do CBR
• Aside: search engines can also solve these problems

Collaborative Filtering approaches
• User based

• Item based

User Similarity
What should we recommend for User 1?

User
2

User
1

Item 1

Item 2

User
3

Item 3

Item 4

User
4

Item Similarity

User
2

User
1

Item 1

Item 2

User
3

Item 3

Item 4

User
4

Component diagram
Core platform DB

QDSS (Cassandra)

DataModel Extractor /
User Preference Builder
Feeder

Recommender App
REST API

Personalization
DB

Key architecture blocks
• Data Extraction
– This involves extracting relevant data from :
• Core Platform DB (MySQL)
• Qyuki custom Data store (Cassandra)

• Data Mining :
– This involves a series of activities to calculate
distances/similarities between items, preferences of users
and applying Machine Learning for further use cases.

• Personalization Service:
– The application of relevant data contextually for the users
and creations on Qyuki, exposed as a REST api.

Entities to Recommend : Creations
For every creation, there are 2 parts to
recommendations :
1) Creations similar to the current creation(ContentBased)
2) Recommendations based on the studied
preferences of the user.

Features cont..
Entities to Recommend : Users
• New users/creators to follow based on his
social graph & other activities
• New users/creators to collaborate with.
– The engine would recommend artists to artists to
collaborate and create.
– QInfluence

Personalized Creations
• Track user’s activities
– MySQL + Custom Analytics(Cassandra)

• Explicit & Implicit Ratings

• Explicit : None in our case

• Implicit : Emotes, comments, views,
engagement with creator

Personalized Creations cont..
• Mahout expects data input in the form :
User
Item
Preference
U253
I306
3.5
U279
I492
4.0
• Preference Compute Engine
– Consider all user activities, infer preferences

• Feed preference data to Mahout, you are done(to some extent)

• Filtering and rescoring

Some Mahout recommendations logic/code

• DataModel : preference data model
– DataModel dataModel = new FileDataModel(new File(preferenceFile));

• Notion of similarity(Distance measure)
– UserSimilarity similarity = new PearsonCorrelationSimilarity(dataModel);

• Neighbourhood(for user based)
– UserNeighborhood neighborhood = NearestNUserNeighborhood(10,
similarity, dataModel);

• Recommender instance
–

GenericUserBasedRecommender userBasedRecommender = new
GenericUserBasedRecommender(dataModel, neighborhood, similarity);

• Recommendations for a user
– topUserBasedRecommendeditems =
userBasedRecommender.recommend(userId, numOfItemsToBeCached, null);

Similarity algorithms
• Pearson correlation–based similarity
– Pearson correlation is a number between –1 and
1
– It measures the tendency of two users’ preference
values to move together—to be relatively high, or
relatively low, on the same items.
– doesn’t take into account the number of items in
which two users’ preferences overlap

Similarity algorithms
• Cosine Similarity
– Ignores 0-0 matches
– A measure of alignment/direction
– Since Cos 0 = 1 (1 means 100% similar)

• Euclidean Distance Similarity
– Based on Euclidean distance between two points
(sq root(sum of squares of diff in coordinates)

Which Similarity to use
• If your data is dense (almost all attributes have
non-zero values) and the magnitude of the
attribute values is important, use distance
measures such as Euclidean.

• If the data is subject to grade-inflation (different
users may be using different scales) use Pearson.
• If the data is sparse consider using Cosine
Similarity.

Hybrid approaches
• Weighted hybrid – Combines scores from each
component using linear formula.
• Switching hybrid – Select one recommender
among candidates.
• Mixed hybrid – Based on the merging and
presentation of multiple ranked lists into one.
– Core algorithm of mixed hybrid merges them into a
single ranked list.

Evaluating Recommenders
•
•
•
•

Mahout
Training Set, Test Set
Iterative preference weight tuning + evaluator
RMSE

• A/B testing
• Offline

Distributed Recommendations with Mahout &
Hadoop

Movie recommendations Case
• Movie rating data (17 million ratings from
around 6000 users available as input)
• Goal : Compute recommendations for users
based on the rating data
• Use Mahout over Hadoop
• Mahout has support to run Map Reduce Jobs
• Runs a series of Map Reduce Jobs to compute
recommendations (Movies a user would like
to watch)
278

Run Mahout job on
• hadoop jar ~/mahout-core-0.7-job.jar
org.apache.mahout.cf.taste.hadoop.item.Reco
mmenderJob --input
/user/hadoop/input/ratings.csv --output
recooutput -s SIMILARITY_COOCCURRENCE
--usersFile /user/hadoop/input/users.txt

Other similarity options
• You can pass one of these as well to the
• “-s” flag :
• SIMILARITY_COOCCURRENCE,
SIMILARITY_EUCLIDEAN_DISTANCE,
SIMILARITY_LOGLIKELIHOOD,
SIMILARITY_PEARSON_CORRELATION,
SIMILARITY_TANIMOTO_COEFFICIENT,
SIMILARITY_UNCENTERED_COSINE,
SIMILARITY_UNCENTERED_ZERO_ASSUMING_COSIN
E
280

Learning

•

Reach out to experienced folks early

•

Recommending follows the 90/10 rule

•

Assign the right task to the right person

•

Use Analytics metrics to evaluate recommenders periodically

•

Importance to human evaluation

•

Test thoroughly

•

Mahout has some bugs(e.g. filtering part)

•

Experiment with hybrid approaches

•

You won’t need Hadoop generally

•

Performance test your application

•

Have a fallback option, hence take charge of your popularity algorithm

What is recommendation?
Recommendation involves the prediction of what
new items a user would like or dislike based on
preferences of or associations to previous items
(Made-up) Example:
A user, John Doe, likes the following books (items):
A Tale of Two Cities
The Great Gatsby
For Whom the Bell Tolls

Recommendations will predict which new books
(items), John Doe, will like:
Jane Eyre
The Adventures of Tom Sawyer
282

How does Mahout’s Recommendation
Engine Work?
5

3

4

4

2

2

1

2

40

3

3

3

2

1

1

0

0

18.5

4

3

4

3

1

2

0

0

24.5

4

2

3

4

2

2

1

2

1

1

2

2

1

1

4.5

26

2

1

2

2

1

2

0

0

16.5

1

0

0

1

1

0

1

5

15.5

U

R

X

S
S is the similarity matrix between items
U is the user’s preferences for items
R is the predicted recommendations

4

=

40

Item 1

Item 2

Item 3

Item 4

Item 5

Item 6

Item 7

What is the similarity matrix, S?

Item 1

5

3

4

4

2

2

1

Item 2

3

3

3

2

2

1

0

Item 3

4

3

4

3

1

2

0

Item 4

4

2

3

4

2

2

1

Item 5

2

2

1

2

2

1

1

Item 6

2

1

2

2

1

2

0

Item 7

1

0

0

1

1

0

 S is a n x n (square) matrix
 Each element, e, in S are indexed
by row (j) and column (k), ejk
 Each ejk in S holds a value that
describes how similar are its
corresponding j-th and k-th items
 In this example, the similarity of
the j-th and k-th items are
determined by frequency of their
co-occurrence (when the j-th item
is seen, the k-th item is seen as
well)
 In general, any similarity measure
may be used to produce these
values

 We see in this example that

1

 Items 1 and 2 co-occur 3 times,
 Items 1 and 3 co-occur 4 times,
 and so on…

S
284

What is the user’s preferences, U?
Item 1
Item 2

0

Item 3

0

Item 4

4

Item 5

4.5

Item 6

0

Item 7

The user’s preference
is represented as a
column vector

2

5

Each value in the vector
represents the user’s
preference for j-th item
In general, this column
vector is sparse
Values of zero, 0,
represent no recorded
preferences for the j-th
item

U

285

What is the recommendation, R?
Item 1

 R is a column vector
representing the prediction
of recommendation of the jth item for the user
 R is computed from the
multiplication of S and U

40

Item 2 18.5
Item 3 24.5
Item 4

40

Item 5

26

SxU=R

 In this running example, the
user already has expressed
positive preferences for Items
1, 4, 5 and 7, so we look at
only Items 2, 3, and 6
 We would recommend to the
user Items 3, 2, and 6, in this
order, to the user

Item 6 16.5
Item 7 15.5
R

286

What data format does Mahout’s
recommendation engine expects?
• For Mahout v0.7, look
at RecommenderJob

Format 1
123,345
123,456
123,789
…
789,458

(org.apache.mahout.cf.taste.hadoop.item.RecommenderJob)

• Each line of the input file
should have the following
format
– userID,itemID[,preferencevalue
]

Format 2

• userID is parsed as a long
• itemID is parsed as a long
• preferencevalue is parsed as a
double and is optional

123,345,1.0
123,456,2.2
123,789,3.4
…
789,458,1.2
287

How do you run Mahout’s
recommendation engine?
 Requirements
Hadoop cluster on GNU/Linux
Java 1.6.x
SSH
 Assuming you have a Hadoop cluster installed and configured
correctly with the data loaded into HDFS,
 hadoop jar ~/mahout-core-0.7-job.jar
org.apache.mahout.cf.taste.hadoop.item.RecommenderJob --input
/user/hadoop/input/ratings.csv --output recooutput -s SIMILARITY_COOCCURRENCE -usersFile /user/hadoop/input/users.txt
 $HADOOP_INSTALL$ is the location where you installed Hadoop
 $TARGET$ is the directory where you have the Mahout jar file
 $INPUT$ is the input file name
 $OUTPUT$ is the output file name

288

Running Mahout RecommenderJob options
•
•
•
•
•
•

•

There are plenty of runtime options (check javadocs)
--userFile (path) : optional; a file containing userIDs; only preferences of these
userIDs will be computed
--itemsFile (path) : optional; a file containing itemIDs; only these items will be used
in the recommendation predictions
--numRecommendations (integer) : number of recommendations to compute per
user; default 10
--booleanData (boolean) : treat input data as having no preference values; default
false
--maxPrefsPerUser (integer) : maximum number of preferences considered per
user in final recommendation phase; default 10
--similarityClassname (classname): similarity measure (cooccurence, euclidean,
log-likelihood, pearson, tanimoto coefficient, uncentered cosine, cosine)

289

Coordination in a distributed system
• Coordination: An act that multiple nodes must
perform together.
• Examples:
– Group membership
– Locking
– Publisher/Subscriber
– Leader Election
– Synchronization

• Getting node coordination correct is very hard!

Introducing ZooKeeper
ZooKeeper allows distributed processes to
coordinate with each other through a shared
hierarchical name space of data registers.
- ZooKeeper Wiki

ZooKeeper is much more than a
distributed lock server!

What is ZooKeeper?
• An open source, high-performance
coordination service for distributed
applications.
• Exposes common services in simple interface:
– naming
– configuration management
– locks & synchronization
– group services
… developers don't have to write them from scratch

• Build your own on it for specific needs.

ZooKeeper Use Cases
• Configuration Management
– Cluster member nodes bootstrapping configuration from a
centralized source in unattended way
– Easier, simpler deployment/provisioning

• Distributed Cluster Management
– Node join / leave
– Node statuses in real time

•
•
•
•

Naming service – e.g. DNS
Distributed synchronization - locks, barriers, queues
Leader election in a distributed system.
Centralized and highly reliable (simple) data registry

The ZooKeeper Service

•
•
•
•
•

ZooKeeper Service is replicated over a set of machines
All machines store a copy of the data (in memory)‫‏‬
A leader is elected on service startup
Clients only connect to a single ZooKeeper server & maintains a TCP connection.
Client can read from any Zookeeper server, writes go through the leader & needs
majority consensus.

Image: https://cwiki.apache.org/confluence/display/ZOOKEEPER/ProjectDescription

The ZooKeeper Data Model
• ZooKeeper has a hierarchal name space.
• Each node in the namespace is called as a ZNode.
• Every ZNode has data (given as byte[]) and can optionally
have children.
parent : "foo"
|-- child1 : "bar"
|-- child2 : "spam"
`-- child3 : "eggs"
`-- grandchild1 : "42"

• ZNode paths:
– canonical, absolute, slash-separated
– no relative references.
– names can have Unicode characters

Hadoop Master Class : A concise overview

Hadoop Master Class : A concise overview

Recomendados

Recomendados

Más contenido relacionado

La actualidad más candente

La actualidad más candente (20)

Similar a Hadoop Master Class : A concise overview

Similar a Hadoop Master Class : A concise overview (20)

Último

Último (20)

Hadoop Master Class : A concise overview