The Hadoop Distributed File System, by Dhyuba Borthakurand Related Work

Presented by Mohit Goenka

The Hadoop Distributed File System: Architecture and

Design

Requirement

SECTION TITLE

• Need to process Multi Petabyte Datasets

• Expensive to build reliability in each application.

• Nodes fail every day• Need common infrastructure

Introduction

SECTION TITLE

• HDFS, Hadoop Distributed File System is designed to run on commodity hardware

• Built out by brilliant engineers and contributors from Yahoo, and Facebook and Cloudera and other companies

• Has grown into really large project at Apache with significant ecosystem

• Typically in 2 level architecture

– Nodes are commodity PCs– 30-40 nodes/rack– Uplink from rack is 3-4 gigabit– Rack-internal is 1 gigabit

Commodity Hardware

SECTION TITLE

Goals

SECTION TITLE

• Very Large Distributed File System– 10K nodes, 100 million files, 10 PB

• Assumes Commodity Hardware– Files are replicated to handle hardware failure– Detect failures and recovers from them

• Optimized for Batch Processing– Data locations exposed so that computations can move to where data resides– Provides very high aggregate bandwidth

• User Space, runs on heterogeneous OS

HDFS Basic Architecture

SECTION TITLESecondaryNameNode

Client

NameNode

DataNodes

1. filename

2. BlckId, DataNodes

o

3.Read data

Cluster Membership

Cluster Membership

NameNode : Maps a file to a file-id and list of MapNodesDataNode : Maps a block-id to a physical location on diskSecondaryNameNode: Periodic merge of Transaction log

Distributed File System

SECTION TITLE

• Single Namespace for entire cluster• Data Coherency

– Write-once-read-many access model– Client can only append to existing files

• Files are broken up into blocks– Typically 128 MB block size– Each block replicated on multiple DataNodes

• Intelligent Client– Client can find location of blocks– Client accesses data directly from DataNode

HDFS Core Architecture

SECTION TITLE

NameNode Metadata

SECTION TITLE

• Meta-data in Memory– The entire metadata is in main memory– No demand paging of meta-data

• Types of Metadata– List of files– List of Blocks for each file– List of DataNodes for each block– File attributes, e.g creation time, replication factor

• A Transaction Log– Records file creations, file deletions. etc

Data Node

SECTION TITLE

• A Block Server– Stores data in the local file system (e.g. ext3)– Stores meta-data of a block (e.g. CRC)– Serves data and meta-data to Clients

• Block Report– Periodically sends a report of all existing blocks to the NameNode

• Facilitates Pipelining of Data– Forwards data to other specified DataNodes

Block Placement

SECTION TITLE

• Current Strategy

- One replica on local node

- Second replica on a remote rack

- Third replica on same remote rack

- Additional replicas are randomly placed• Clients read from nearest replica• Would like to make this policy

pluggable

Data Correctness

SECTION TITLE

• Use Checksums to validate data

– Use CRC32• File Creation

– Client computes checksum per 512 byte

– DataNode stores the checksum• File access

– Client retrieves the data and checksum from DataNode

– If Validation fails, Client tries other replicas

NameNode Failure

SECTION TITLE

• A single point of failure• Transaction Log stored in multiple

directories

- A directory on the local file system

- A directory on a remote file system (NFS/CIFS)

• Need to develop a real HA solution

Data Pipelining

SECTION TITLE

• Client retrieves a list of DataNodes on which to place replicas of a block

• Client writes block to the first DataNode

• The first DataNode forwards the data to the next DataNode in the Pipeline

• When all replicas are written, the Client moves on to write the next block in file

Rebalancer

SECTION TITLE

• Goal: % disk full on DataNodes should be similar– Usually run when new DataNodes are

added– Cluster is online when Rebalancer is

active– Rebalancer is throttled to avoid

network congestion– Command line tool

Hadoop Map / Reduce

SECTION TITLE

• The Map-Reduce programming model– Framework for distributed processing of large data sets– Pluggable user code runs in generic framework

• Common design pattern in data processing cat * | grep | sort | unique -c | cat > file

input | map | shuffle | reduce | output• Natural for:

– Log processing – Web search indexing – Ad-hoc queries

Data Flow

SECTION TITLEWeb Servers Scribe Servers

Network Storage

Hadoop ClusterOracle RAC MySQL

Basic Operations

SECTION TITLE

• Listing files- ./bin/hadoop fs –ls

• Writing files- ./bin/hadoop fs –put

• Running Map Reduce Jobs- mkdir input - cp conf/*.xml input - cat output/*

Hadoop Ecosystem Projects

SECTION TITLE

• HBase- Big Table

• HIVE- Built on Facebook, provides SQL

interface

• Chukwa- Log Processing

• Pig- Scientific data analysis language

• Zookeeper- Distributed Systems management

Limitatons

SECTION TITLE

• The gigabytes to terabytes of data this system handles can only be scaled down to limited threshold

• Due to this threshold being very high, the system is limited in a lot of ways

• It hampers the efficiency of the system during large computations or parallel data exchange

JSON Interface to Control HDFSAn Open Source Project

by Mohit Goenka

JSON Interface to Control HDFS

An Open Source Project by Mohit Goenka

JSON

SECTION TITLE

• JSON (JavaScript Object Notation) is a lightweight data-interchange format

• Can be easily read and written by humans

• Can be easily parsed by machines

• Written in text format• Similar conventions as existing

programming languages

JSON Data

SECTION TITLE

• It is based on two structures:- A collection of name/value

pairs- An ordered list of values

• Concept: Use the light-weighted nature of JSON data to automate command execution on HDFS interface

Goal

SECTION TITLE

• Designing a JSON interface to control HDFS

• Development of two modules:- For writing into the system- For reading from the system

Outcome

SECTION TITLE

• User can specify execution commands directly in the JSON file along with data

• Only data gets stored into the system

• Commands are deleted from the file after execution

Sources and Acknowledgements

Sources

SECTION TITLE

• Dhurba Borthakur, Apache Hadoop Developer, Facebook Data Infrastructure

• Matei Zaharia, Cloudera / Facebook / UC Berkeley RAD Lab

• Devaraj Das, Yahoo! Inc. Bangalore and Apache Software Foundation

• HDFS Java API: - http://hadoop.apache.org/core/docs/current/api/

• HDFS source code: - http://hadoop.apache.org/core/version_control.html

Acknowledgements

SECTION TITLE

• Professor Chris Mattmann for guidance as and when reqired

• Hossein (Farshad) Tajalli for his continued support and help throughout the project

• All my classmates for providing valuable inputs throughout the work, especially through their presentations

That’s All Folks!

SECTION TITLE