A new way to store and analyze data

Sandesh Deshmane

• What is Hadoop?

• Why, Where, When?

• Benefits of Hadoop

• How Hadoop Works?

• Hadoop Architecture

• HDFS

• Hadoop MapReduce

• Installation & Execution

• Demo

Topics Covered

In pioneer days they used oxen for heavy pulling, and when one ox couldn’t budge a log, they didn’t try to grow a larger ox. We shouldn’t be trying for bigger computers, but more systems of computers.

—Grace Hopper

History

• Size of digital Universe was estimated 0.18 zeta byte in 2006 and was 3 zeta byte in 2012

1 zeta byte =10^21 bytes=1k exa bytes=1 million petabyte=

1 billion terabytes

• The New York stock exchange generate 1TB data per day

• Facebook stores around 10 billion photos . Around 1 petabyte.

• The internet archive stores 1 peta byte data and its growing ( 20 TB per month).

Background

• Created by Douglas Reed Cutting,

• Open-source Apache Software Foundation.

• consists of two key services:

a. Reliable data storage using the Hadoop Distributed File System (HDFS).b. High-performance parallel data processing using a technique called Map Reduce.

• Hadoop is large-scale, high-performance processing jobs — in spite of system changes or failures.

What is Hadoop?

• Need to process 100TB datasets

• On 1 node:– scanning @ 50MB/s = 23 days

• On 1000 node cluster:– scanning @ 50MB/s = 33 min

• Need Efficient, Reliable and Usable framework

Hadoop, Why?

Where• Batch data processing, not

real-time / user facing (e.g. Document Analysis and Indexing, Web Graphs and Crawling)

• Highly parallel data intensive distributed applications

• Very large production deployments

When

• Process lots of unstructured data

• When your processing can easily be made parallel

• Running batch jobs is acceptable

• When you have access to lots of cheap hardware

Where and When Hadoop?

• Runs on cheap commodity hardware

• Automatically handles data replication and node failure

• It does the hard work – you can focus on processing data

• Cost Saving and efficient and reliable data processing

Benefits of Hadoop

• Hadoop implements a computational paradigm named Map/Reduce, where the application is divided into many small fragments of work, each of which may be executed or re-executed on any node in the cluster.

• In addition, it provides a distributed file system (HDFS) that stores data on the compute nodes, providing very high aggregate bandwidth across the cluster.

• Both Map/Reduce and the distributed file system are designed so that node failures are automatically handled by the framework.

How Hadoop Works?

The Apache Hadoop project develops open-source software for reliable, scalable, distributed computing

Hadoop Consists of:

• Hadoop Common: The common utilities that support the other Hadoop subprojects.

• HDFS: A distributed file system that provides high throughput access to application data.

• MapReduce: A software framework for distributed processing of large data sets on compute clusters.

Hadoop Architecture

Web Servers

Scribe Servers

Network Storage

Hadoop ClusterOracle DB MySQL

Hadoop Architecture

• Java

• Python

• Ruby

• C++ (Hadoop Pipes)

Supported Languages

• Know as Hadoop Distribute File System

• Primary storage system for Hadoop Apps

• Multiple replicas of data blocks distributed on compute nodes for reliability

• Files are stored on multiple boxes for durability and high availability

HDFS

• Distributed File System = holds large amount of data and provides access to this data to many clients distributed across a network . e.g NFS

• HDFS stores large amount of Information than DFS

• HDFS stores data reliably.

• HDFS provides fast, scalable access to this information to large number of clients in Cluster

DFS vs. HDFS

• Optimized for long sequential reads

• Data written once , read multiple times, no append possible

• Large file, sequential reads so no local caching of data.

• Data replication

HDFS

HDFS Architecture

• Block Structure files system

• File is divided to bocks and stored

• Each individual machine in cluster is Data Node

• Default block size is 64 MB

• Information of blocks is stored in metadata

• All this meta data is stored on machine which is Name Node

HDFS Architecture

Data Node and Data Name

<configuration>

<property>

<name>fs.default.name</name>

<value>hdfs://your.server.name.com:9000</value>

</property>

<property>

<name>dfs.data.dir</name>

<value>/home/username/hdfs/data</value>

</property>

<property>

<name>dfs.name.dir</name>

<value>/home/username/hdfs/name</value>

</property>

</configuration>

HDFS Config File

public class HDFSHelloWorld {

public static final String theFilename = "hello.txt";

public static final String message = "Hello, world!\n";

public static void main (String [] args) throws IOException {

Configuration conf = new Configuration();

FileSystem fs = FileSystem.get(conf);

Path filenamePath = new Path(theFilename);

try {

if (fs.exists(filenamePath)) {

// remove the file first

fs.delete(filenamePath);

}

FSDataOutputStream out = fs.create(filenamePath);

out.writeUTF(message);

out.close();

FSDataInputStream in = fs.open(filenamePath);

String messageIn = in.readUTF();

System.out.print(messageIn);

in.close();

} catch (IOException ioe) {

System.err.println("IOException during operation: " + ioe.toString());

System.exit(1);

}

}

Sample Java Code to Read/Write from HDFS

Map Reduce

Cluster Look

Map

Reduce

• HDFS handles the Distributed File System layer

• MapReduce is how we process the data

• MapReduce Daemons

JobTracker

TaskTracker

• Goals

Distribute the reading and processing of data

Localize the processing when possible

Share as little data as possible while processing

MapReduce

MapReduce

• One per cluster “master node”

• Takes jobs from clients

• Splits work into “tasks”

• Distributes “tasks” to TaskTrackers

• Monitors progress, deals with failures

Job Tracker

• Many per cluster “slave nodes”

• Does the actual work, executes the code for the job

• Talks regularly with JobTracker

• Launches child process when given a task

• Reports progress of running “task” back to JobTracker

Task Tracker

• Client Submits job:I want to count the count of each word We will assume that the data to process is already there in HDFS

• Job Tracker receives job• Queries the NamNode for number of blocks in File• The job is split into Tasks• One map task per each block• As many reduce tasks as specified in the Job

• TaskTracker checks in Regularly with JobTracker Is there any work for me ?• If the JobTracker has a MapTask that the TaskTracker has a local block for the file being processed then the TaskTracker will be given

the “task”

Anatomy of Map Reduce Job

Map Reduce Job – Big Picture

Client Submits to JobTracker

JobTracker Queries Name Node for Block Info

Job tracker Defines Job as Collection of Tasks

Task Trackers Checking in are Assigned tasks

Task Trackers Checking in are Assigned tasks

• Read text files and count how often words occur. o The input is text fileso The output is a text file

each line: word, tab, count

• Map: Produce pairs of (word, count)

• Reduce: For each word, sum up the counts.

Example of MapReduce - Word Count

public static class MapClass extends MapReduceBase implements Mapper<LongWritable, Text, Text, IntWritable>

{

private final static IntWritable one = new IntWritable(1);

private Text word = new Text();

public void map (LongWritable key, Text value, OutputCollector<Text, IntWritable> output, Reporter reporter) throws IOException

{

String line = value.toString();

StringTokenizer itr = new StringTokenizer(line);

while (itr.hasMoreTokens())

{

word.set(itr.nextToken());

output.collect(word, one);

}

}

}

Map Class

public static class ReduceClass extends MapReduceBase implements Reducer<Text, IntWritable, Text, IntWritable>

{

public void reduce(Text key, Iterator<IntWritable> values, OutputCollector<Text, IntWritable> output, Reporter reporter) throws IOException

{

int sum = 0;

while (values.hasNext())

{

sum += values.next().get();

}

output.collect(key, new IntWritable(sum));

}

}

Reduce Class

public void run(String inputPath, String outputPath) throws Exception

{

JobConf conf = new JobConf(WordCount.class); conf.setJobName("wordcount"); // the keys are words (strings)

conf.setOutputKeyClass(Text.class); // the values are counts (ints) conf.setOutputValueClass(IntWritable.class);

conf.setMapperClass(MapClass.class); conf.setReducerClass(Reduce.class);

FileInputFormat.addInputPath(conf, new Path(inputPath)); FileOutputFormat.setOutputPath(conf, new Path(outputPath)); JobClient.runJob(conf);

}

Driver Class

import static org.mockito.Matchers.anyObject;

import static org.mockito.Mockito.*;

import java.io.IOException;

import org.apache.hadoop.io.*;

import org.apache.hadoop.mapred.OutputCollector;

import org.junit.*;

public class WordCountMapperTest {

@Test

public void processesValidRecord() throws IOException {

MapClass mapper = new MapClass ();

Text value = new Text(“test test”)

OutputCollector<Text, IntWritable> output = mock(OutputCollector.class);

mapper.map(null, value, output, null);

verify(output).collect(new Text(“test"), new IntWritable(2));

}

}

Junit For Mapper

Junit for Reducer

import static org.mockito.Matchers.anyObject;

import static org.mockito.Mockito.*;

import java.io.IOException;

import org.apache.hadoop.io.*;

import org.apache.hadoop.mapred.OutputCollector;

import org.junit.*;

@Test

public void returnsMaximumIntegerInValues() throws IOException {

ReduceClass reducer = new ReduceClass ();

Text key = new Text(“test");

Iterator<IntWritable> values = Arrays.asList(

new IntWritable(1), new IntWritable(1)).iterator();

OutputCollector<Text, IntWritable> output = mock(OutputCollector.class);

reducer.reduce(key, values, output, null);

verify(output).collect(key, new IntWritable(2));

}

Installation :• Requirements: Linux, Java 1.6,

sshd,• Configure SSH for password-free

authentication• Unpack Hadoop distribution• Edit a few configuration files• Format the DFS on the name

node• Start all the daemon processes

Execution:• Compile your job into a JAR file• Copy input data into HDFS• Execute bin/hadoop jar with

relevant args• Monitor tasks via Web interface

(optional)• Examine output when job is

complete

Let’s Go…

Demo

Hadoop Users

• Adobe• Alibaba• Amazon• AOL• Facebook• Google• IBM

Major Contributor

• Apache• Cloudera• Yahoo

Hadoop Community

• Apache Hadoop! (http://hadoop.apache.org )• Hadoop on Wikipedia (

http://en.wikipedia.org/wiki/Hadoop)• Free Search by Doug Cutting (

http://cutting.wordpress.com )• Hadoop and Distributed Computing at Yahoo! (

http://developer.yahoo.com/hadoop )• Cloudera - Apache Hadoop for the Enterprise (

http://www.cloudera.com )

References