Prepared By : Manoj Kumar Joshi & Vikas Sawhney

General Agenda

Introduction to Hadoop Architecture

 Acknowledgement
Thanks to all the authors who left their self-
explanatory images on the internet.

Thanks to bradhedlund.com and Cloudera Inc for

their blogs on Hadoop

We own the errors of course

 History Of Hadoop
 Hadoop was started by Doug Cutting to support two of his other
well known projects, Lucene and Nutch

 Hadoop has been inspired by Google's File System (GFS) which

was detailed in a paper by released by Google in 2003

 Hadoop, originally called Nutch Distributed File System (NDFS)

split from Nutch in 2006 to become a sub-project of Lucene. At
this point it was renamed to Hadoop.
History Of Hadoop (Contd.)
 Distributed File System (DFS)
 A Distributed File System ( DFS ) is simply a classical model of
a file system distributed across multiple machines. The purpose is
to promote sharing of dispersed files.

 This is an area of active research interest today.

 Clients should view a DFS the same way they would a centralized
FS; the distribution is hidden at a lower level.

 A DFS provides high throughput data access and fault tolerance.

 Why Hadoop ?
Hadoop infrastructure provides these capabilities
 Scalability
-Thousands of Compute Nodes
-Petabytes of data
 Cost effective
-Runs On Low Cost Commodity Hardware
 Efficient
-By distributing the data, Hadoop can process it in
parallel on the nodes where the data is located.
 What Is Hadoop ?
 Open source software platform for scalable, distributed computing

 Hadoop provides fast and reliable analysis of both structured data

and unstructured data

 Apache Hadoop software library is essentially a framework that

allows for the distributed processing of large datasets across
clusters of computers using a simple programming model.

 Hadoop can scale up from single servers to thousands of

machines, each offering local computation and storage.
Hadoop Architecture
Hadoop Stack
 Hadoop Ecosystems Projects
 Hadoop Ecosystem Projects includes:
 Hadoop Common utilities
 Avro: A data serialization system with scripting languages.
 Chukwa: managing large distributed systems.
 HBase: A scalable, distributed database for large tables.
 HDFS: A distributed file system.
 Hive: data summarization and ad hoc querying.
 MapReduce: distributed processing on compute clusters.
 Pig: A high-level data-flow language for parallel computation.
 ZooKeeper: coordination service for distributed applications.
 Use Cases for Hadoop
 To aggregate “data exhaust” — messages, posts, blog entries,
photos, video clips, maps, web graph

 To give data context — friends networks, social graphs,

recommendations, collaborative filtering

 To keep apps running — web logs, system logs, system metrics,

database query logs

 To deliver novel mashup services – mobile location data,

clickstream data, SKUs, pricing
Hadoop Server Roles
Assumptions and Goals of HDFS
 Hardware Failure

 Streaming Data Access (Best for batch processing)

 Large Data Sets

 Simple Coherency Model (write-once-read-many access model)

 Portability Across Heterogeneous Hardware and Software Platforms

 HDFS (Hadoop Distributed File
System)
 A distributed file system that provides high-throughput access to
application data

 HDFS uses a master/slave architecture in which one device

(master) termed as NameNode controls one or more other devices
(slaves) termed as DataNode.

 It breaks Data/Files into small blocks (128 MB each block) and

stores on DataNode and each block replicates on other nodes to
accomplish fault tolerance.

 NameNode keeps the track of blocks written to the DataNode.

HDFS Cluster Architecture
HDFS Architecture
HDFS Cluster Architecture
 HDFS Daemons
 NAME NODE

 DATA NODE

 SECONDRY NAME NODE

 Name Node
 Keeps the metadata of all files/blocks in the file system, and
tracks where across the cluster the file data is kept. It does not
store the data of these files itself. Kind of block lookup dictionary(
index or address book of blocks).

 Client applications talk to the NameNode whenever they wish to

locate a file, or when they want to add/copy/move/delete a file. The
NameNode responds the successful requests by returning a list of
relevant DataNode servers where the data lives
Name Node (Contd.)
 Name Node (Contd.)
fsimage - Its the snapshot of the filesystem when NameNode started

Edit logs - Its the sequence of changes made to the filesystem after NameNode started
 Data Node
 DataNode stores data in the Hadoop Filesystem

 A functional filesystem has more than one DataNode, with data

replicated across them

 On startup, a DataNode connects to the NameNode; spinning

until that service comes up. It then responds to requests from the
NameNode for filesystem operations.

 Client applications can talk directly to a DataNode, once the

NameNode has provided the location of the data
 Data Replication
 Why need data replication ?

 HDFS is designed to handle large scale data in distributed environment

 Hardware or software failure, or network partition exist
 Therefore need replications for those fault tolerance
 Replication (Contd.)
 Replication factor
 Decided by users, and can be dynamically tuned.

 How to Create replications efficiently ?

 Replication pipeline: Instead of single machine create replications, a
pipe line is applied
 Machine 1 make replication to machine 2, at the same time machine 2
make the replication to machine 3, etc.

 Replication placement
 High initialization time to create replication to all machines
 An approximate solution: Only 3 replications
One replication resides in current node
One replication resides in current rack
One replication resides in another rack
Replication Pipeline
Rack Awareness
 Data Node Failure
Data Node Failure Condition
If a data node failed, Name Node could know the blocks it
contains, create same replications to other alive nodes, and
unregister this dead node

Data Integrity
Corruption may occur in network transfer, Hardware failure etc.
Apply checksum checking on the contents of files on HDFS, and
store the checksum in HDFS namespace
If checksum is not correct after fetching, drop it and fetch another
replication from other machines.
Heartbeats and Re-Replication
 Secondary Name Node
 Not a failover NameNode

 The only purpose of the secondary name-node is to perform

periodic checkpoints. The secondary name-node periodically
downloads current name-node image and edits log files, joins
them into new image and uploads the new image back to the
(primary and the only) name-node

 Default checkpoint time is one hour. It can be set to one minute

on highly busy clusters where lots of write operations are being
performed.
Secondary Name Node (Contd.)
 Name Node Failure
 NameNode is the single point of failure in the cluster

 If NameNode is down due to software glitch, restart the machine

 If original NameNode can be restored, secondary can re-establish

the most current metadata snapshot

 If machine don’t come up, metadata for the cluster is

irretrievable. In this situation create a new NameNode, use
secondary to copy metadata to new primary, restart whole cluster

 Trick : Bring new NameNode up, but use DNS to make cluster
believe it’s the original one
 Apache MapReduce
 A software framework for distributed processing of large data sets

 The framework takes care of scheduling tasks, monitoring them

and re-executing any failed tasks.

 It splits the input data set into independent chunks that are
processed in a completely parallel manner.

 MapReduce framework sorts the outputs of the maps, which are

then input to the reduce tasks. Typically, both the input and the
output of the job are stored in a file system.
MapReduce Architecture
Map Task
Reduce Task
 MapReduce Dataflow
 An input reader
 A Map function
 A partition function
 A compare function
 A Reduce function
 An output writer
 MapReduce Daemons
 JOB TRACKER

 TASK TRACKER
 JobTracker
 JobTracker is the daemon service for submitting and tracking
MapReduce jobs in Hadoop

 JobTracker performs following actions in Hadoop :

 It accepts the MapReduce Jobs from client applications

 Talks to NameNode to determine data location
 Locates available TaskTracker Node
 Submits the work to the chosen TaskTracker Node
 TaskTracker
 A TaskTracker node accepts map, reduce or shuffle operations
from a JobTracker

 Its configured with a set of slots, these indicate the number of

tasks that it can accept

 JobTracker seeks for the free slot to assign a job

 TaskTracker notifies the JobTracker about job success status.

 TaskTracker also sends the heartbeat signals to the job tracker to

ensure its availability, it also reports the no. of available free slots
with it.
 Hadoop Configuration
 Untar hadoop-*.**.*.tar.gz to your user path
About Version:
The latest stable version 1.0.4 is recommended.
 edit the file conf/hadoop-env.sh to define at least JAVA_HOME
to be the root of your Java installation

 edit the files to configure properties:

conf/core-site.xml: conf/hdfs-site.xml: conf/mapred-site.xml:
<configuration> <configuration> <configuration>
<property> <property> <property>
<name> <name> <name>
fs.default.name dfs.replication mapred.job.tracker
</name> </name> </name>
<value> <value> <value>
hdfs://localhost:9000 1 localhost:9001
</value> </value> </value>
</property> </property> </property>
</configuration> </configuration> </configuration>
 Hadoop Releases
 1.0.X - current stable version, 1.0 release
 1.1.X - current beta version, 1.1 release
 2.X.X - current alpha version
 0.23.X - simmilar to 2.X.X but missing NN HA.
 0.22.X - does not include security
 0.20.203.X - old legacy stable version
 0.20.X - old legacy version
 HDFS Access Methods
 Java API (For applications)

 Browser Interface (Next Slide)

 Hadoop FS Shell
 Formatting filesystem with HDFS
 bin/hadoop namenode -format
 To add a directory
 bin/hadoop dfs –mkdir abc

 To list a directory
 bin/hadoop dfs -ls /

 To display content of a file

 bin/hadoop dfs -cat filename
 Hadoop Browser Interfaces
 MapReduce Job Tracker Web Interface
http://localhost:50030/
 Task Tracker Web Interface
http://localhost:50060/
 HDFS Name Node Web Interface
http://localhost:50070/
Name Node Interface
 References
 http://hadoop.apache.org/
http://wiki.apache.org/hadoop/
http://hadoop.apache.org/core/docs/current/hdfs_des
ign.html
http://wiki.apache.org/hadoop/FAQ
 Questions?
 Write your queries to [email protected]