This document provides an overview of Hadoop architecture. It discusses how Hadoop uses MapReduce and HDFS to process and store large datasets reliably across commodity hardware. MapReduce allows distributed processing of data through mapping and reducing functions. HDFS provides a distributed file system that stores data reliably in blocks across nodes. The document outlines components like the NameNode, DataNodes and how Hadoop handles failures transparently at scale.

1. Hadoop Overview &
Architecture
Milind Bhandarkar
Chief Scientist, Machine Learning Platforms,
Greenplum, A Division of EMC
(Twitter: @techmilind)

2. About Me
• http://www.linkedin.com/in/milindb
• Founding member of Hadoop team at Yahoo! [2005-2010]
• Contributor to Apache Hadoop since v0.1
• Built and led Grid Solutions Team at Yahoo! [2007-2010]
• Parallel Programming Paradigms [1989-today] (PhD cs.illinois.edu)
• Center for Development of Advanced Computing (C-DAC), National Center
for Supercomputing Applications (NCSA), Center for Simulation of Advanced
Rockets, Siebel Systems, Pathscale Inc. (acquired by QLogic), Yahoo!, LinkedIn,
and EMC-Greenplum

3. Agenda
• Motivation
• Hadoop
• Map-Reduce
• Distributed File System
• Hadoop Architecture
• Next Generation MapReduce
• Q & A
2

4. Hadoop At Scale
(Some Statistics)
• 40,000 + machines in 20+ clusters
• Largest cluster is 4,000 machines
• 170 Petabytes of storage
• 1000+ users
• 1,000,000+ jobs/month
3

5. BEHIND
EVERY CLICK

6. Hadoop Workflow

7. Who Uses Hadoop ?

8. Why Hadoop ?
7

9. Big Datasets
(Data-Rich Computing theme proposal. J. Campbell, et al, 2007)

10. Cost Per Gigabyte
(http://www.mkomo.com/cost-per-gigabyte)

11. Storage Trends
(Graph by Adam Leventhal, ACM Queue, Dec 2009)

12. Motivating Examples
11

13. Yahoo! Search Assist

14. Search Assist
• Insight: Related concepts appear close
together in text corpus
• Input: Web pages
• 1 Billion Pages, 10K bytes each
• 10 TB of input data
• Output: List(word, List(related words))
13

15. Search Assist
// Input: List(URL, Text)
foreach URL in Input :
Words = Tokenize(Text(URL));
foreach word in Tokens :
Insert (word, Next(word, Tokens)) in Pairs;
Insert (word, Previous(word, Tokens)) in Pairs;
// Result: Pairs = List (word, RelatedWord)
Group Pairs by word;
// Code Samples
// Result: List (word, List(RelatedWords)
foreach word in Pairs :
Count RelatedWords in GroupedPairs;
// Result: List (word, List(RelatedWords, count))
foreach word in CountedPairs :
Sort Pairs(word, *) descending by count;
choose Top 5 Pairs;
// Result: List (word, Top5(RelatedWords))
14

16. People You May Know

17. People You May Know
• Insight: You might also know Joe Smith if a
lot of folks you know, know Joe Smith
• if you don t know Joe Smith already
• Numbers:
• 100 MM users
• Average connections per user is 100
16

18. People You May Know
Input: List(UserName, List(Connections))
//
foreach u in UserList : // 100 MM
foreach x in Connections(u) : // 100
// Code foreach y in Connections(x) : // 100
Samples
if (y not in Connections(u)) :
Count(u, y)++; // 1 Trillion Iterations
Sort (u,y) in descending order of Count(u,y);
Choose Top 3 y;
Store (u, {y0, y1, y2}) for serving;
17

19. Performance
• 101 Random accesses for each user
• Assume 1 ms per random access
• 100 ms per user
• 100 MM users
• 100 days on a single machine
18

20. MapReduce Paradigm
19

21. Map Reduce
• Primitives in Lisp ( Other functional
languages) 1970s
• Google Paper 2004
• http://labs.google.com/papers/
mapreduce.html
20

22. Map
Output_List = Map (Input_List)
Square (1, 2, 3, 4, 5, 6, 7, 8, 9, 10) =
(1, 4, 9, 16, 25, 36,49, 64, 81, 100)
21

23. Reduce
Output_Element = Reduce (Input_List)
Sum (1, 4, 9, 16, 25, 36,49, 64, 81, 100) = 385
22

24. Parallelism
• Map is inherently parallel
• Each list element processed independently
• Reduce is inherently sequential
• Unless processing multiple lists
• Grouping to produce multiple lists
23

25. Search Assist Map
// Input: http://hadoop.apache.org
Pairs = Tokenize_And_Pair ( Text ( Input ) )
// Example
24

26. Search Assist Reduce
// Input: GroupedList (word, GroupedList(words))
CountedPairs = CountOccurrences (word, RelatedWords)
Output = {
(hadoop, apache, 7) (hadoop, DFS, 3) (hadoop, streaming,
4) (hadoop, mapreduce, 9) ...
}
25

27. Issues with Large Data
• Map Parallelism: Chunking input data
• Reduce Parallelism: Grouping related data
• Dealing with failures load imbalance
26

29. Apache Hadoop
• January 2006: Subproject of Lucene
• January 2008: Top-level Apache project
• Stable Version: 1.0.3
• Latest Version: 2.0.0 (Alpha)
28

30. Apache Hadoop
• Reliable, Performant Distributed file system
• MapReduce Programming framework
• Ecosystem: HBase, Hive, Pig, Howl, Oozie,
Zookeeper, Chukwa, Mahout, Cascading,
Scribe, Cassandra, Hypertable, Voldemort,
Azkaban, Sqoop, Flume, Avro ...
29

31. Problem: Bandwidth to
Data
• Scan 100TB Datasets on 1000 node cluster
• Remote storage @ 10MB/s = 165 mins
• Local storage @ 50-200MB/s = 33-8 mins
• Moving computation is more efficient than
moving data
• Need visibility into data placement
30

32. Problem: Scaling Reliably
• Failure is not an option, it s a rule !
• 1000 nodes, MTBF 1 day
• 4000 disks, 8000 cores, 25 switches, 1000
NICs, 2000 DIMMS (16TB RAM)
• Need fault tolerant store with reasonable
availability guarantees
• Handle hardware faults transparently
31

33. Hadoop Goals
• Scalable: Petabytes (10 15 Bytes) of data on
thousands on nodes
• Economical: Commodity components only
• Reliable
• Engineering reliability into every application
is expensive
32

34. Hadoop MapReduce
33

35. Think MapReduce
• Record = (Key, Value)
• Key : Comparable, Serializable
• Value: Serializable
• Input, Map, Shuffle, Reduce, Output
34

36. Seems Familiar ?
cat /var/log/auth.log* |
grep “session opened” | cut -d’ ‘ -f10 |
sort | Samples
// Code
uniq -c
~/userlist
35

37. Map
• Input: (Key , Value )
1 1
• Output: List(Key , Value )
2 2
• Projections, Filtering, Transformation
36

38. Shuffle
• Input: List(Key , Value )
2 2
• Output
• Sort(Partition(List(Key , List(Value ))))
2 2
• Provided by Hadoop
37

39. Reduce
• Input: List(Key , List(Value ))
2 2
• Output: List(Key , Value )
3 3
• Aggregation
38

40. Hadoop Streaming
• Hadoop is written in Java
• Java MapReduce code is native
• What about Non-Java Programmers ?
• Perl, Python, Shell, R
• grep, sed, awk, uniq as Mappers/Reducers
• Text Input and Output
39

41. Hadoop Streaming
• Thin Java wrapper for Map Reduce Tasks
• Forks actual Mapper Reducer
• IPC via stdin, stdout, stderr
• Key.toString() t Value.toString() n
• Slower than Java programs
• Allows for quick prototyping / debugging
40

42. Hadoop Streaming
bin/hadoop jar hadoop-streaming.jar
$
-input in-files -output out-dir
-mapper mapper.sh -reducer reducer.sh
mapper.sh
#
// Code Samples
sed -e 's/ /n/g' | grep .
#
reducer.sh
uniq -c | awk '{print $2 t $1}'
41

43. Hadoop Distributed File
System (HDFS)
42

44. HDFS
• Data is organized into files and directories
• Files are divided into uniform sized blocks
(default 128MB) and distributed across
cluster nodes
• HDFS exposes block placement so that
computation can be migrated to data
43

45. HDFS
• Blocks are replicated (default 3) to handle
hardware failure
• Replication for performance and fault
tolerance (Rack-Aware placement)
• HDFS keeps checksums of data for
corruption detection and recovery
44

46. HDFS
• Master-Worker Architecture
• Single NameNode
• Many (Thousands) DataNodes
45

47. HDFS Master
(NameNode)
• Manages filesystem namespace
• File metadata (i.e. inode )
• Mapping inode to list of blocks + locations
• Authorization Authentication
• Checkpoint journal namespace changes
46

48. Namenode
• Mapping of datanode to list of blocks
• Monitor datanode health
• Replicate missing blocks
• Keeps ALL namespace in memory
• 60M objects (File/Block) in 16GB
47

49. Datanodes
• Handle block storage on multiple volumes
block integrity
• Clients access the blocks directly from data
nodes
• Periodically send heartbeats and block
reports to Namenode
• Blocks are stored as underlying OS s files
48

50. HDFS Architecture

51. Example: Unigrams
• Input: Huge text corpus
• Wikipedia Articles (40GB uncompressed)
• Output: List of words sorted in descending
order of frequency

52. Unigrams
$ cat ~/wikipedia.txt |
sed -e 's/ /n/g' | grep . |
sort |
uniq -c
~/frequencies.txt
$ cat ~/frequencies.txt |
# cat |
sort -n -k1,1 -r |
# cat
~/unigrams.txt

53. MR for Unigrams
mapper (filename, file-contents):
for each word in file-contents:
emit (word, 1)
reducer (word, values):
sum = 0
for each value in values:
sum = sum + value
emit (word, sum)

54. MR for Unigrams
mapper (word, frequency):
emit (frequency, word)
reducer (frequency, words):
for each word in words:
emit (word, frequency)

55. Unigrams: Java Mapper
public static class MapClass extends MapReduceBase
implements MapperLongWritable, Text, Text, IntWritable {
public void map(LongWritable key, Text value,
OutputCollectorText, IntWritable output,
Reporter reporter) throws IOException {
String line = value.toString();
StringTokenizer itr = new StringTokenizer(line);
while (itr.hasMoreTokens()) {
Text word = new Text(itr.nextToken());
output.collect(word, new IntWritable(1));
}
}
}

56. Unigrams: Java Reducer
public static class Reduce extends MapReduceBase
implements ReducerText, IntWritable, Text, IntWritable {
public void reduce(Text key,IteratorIntWritable values,
OutputCollectorText,IntWritable output,
Reporter reporter) throws IOException {
int sum = 0;
while (values.hasNext()) {
sum += values.next().get();
}
output.collect(key, new IntWritable(sum));
}
}

57. Unigrams: Driver
public void run(String inputPath, String outputPath) throws
Exception
{
JobConf conf = new JobConf(WordCount.class);
conf.setJobName(wordcount);
conf.setMapperClass(MapClass.class);
conf.setReducerClass(Reduce.class);
FileInputFormat.addInputPath(conf, new Path(inputPath));
FileOutputFormat.setOutputPath(conf, new Path(outputPath));
JobClient.runJob(conf);
}

58. Configuration
• Unified Mechanism for
• Configuring Daemons
• Runtime environment for Jobs/Tasks
• Defaults: *-default.xml
• Site-Specific: *-site.xml
• final parameters

59. Example
configuration
property
namemapred.job.tracker/name
valuehead.server.node.com:9001/value
/property
property
namefs.default.name/name
valuehdfs://head.server.node.com:9000/value
/property
property
namemapred.child.java.opts/name
value-Xmx512m/value
finaltrue/final
/property
....
/configuration

60. Running Hadoop Jobs

61. Running a Job
[milindb@gateway ~]$ hadoop jar
$HADOOP_HOME/hadoop-examples.jar wordcount
/data/newsarchive/20080923 /tmp/
newsoutinput.FileInputFormat: Total input paths to
process : 4
mapred.JobClient: Running job: job_200904270516_5709
mapred.JobClient: map 0% reduce 0%
mapred.JobClient: map 3% reduce 0%
mapred.JobClient: map 7% reduce 0%
....
mapred.JobClient: map 100% reduce 21%
mapred.JobClient: map 100% reduce 31%
mapred.JobClient: map 100% reduce 33%
mapred.JobClient: map 100% reduce 66%
mapred.JobClient: map 100% reduce 100%
mapred.JobClient: Job complete: job_200904270516_5709

62. Running a Job
mapred.JobClient: Counters: 18
mapred.JobClient: Job Counters
mapred.JobClient: Launched reduce tasks=1
mapred.JobClient: Rack-local map tasks=10
mapred.JobClient: Launched map tasks=25
mapred.JobClient: Data-local map tasks=1
mapred.JobClient: FileSystemCounters
mapred.JobClient: FILE_BYTES_READ=491145085
mapred.JobClient: HDFS_BYTES_READ=3068106537
mapred.JobClient: FILE_BYTES_WRITTEN=724733409
mapred.JobClient: HDFS_BYTES_WRITTEN=377464307

63. Running a Job
mapred.JobClient: Map-Reduce Framework
mapred.JobClient: Combine output records=73828180
mapred.JobClient: Map input records=36079096
mapred.JobClient: Reduce shuffle bytes=233587524
mapred.JobClient: Spilled Records=78177976
mapred.JobClient: Map output bytes=4278663275
mapred.JobClient: Combine input records=371084796
mapred.JobClient: Map output records=313041519
mapred.JobClient: Reduce input records=15784903

64. JobTracker WebUI
// Code Samples

65. JobTracker Status

66. Jobs Status

67. Job Details

68. Job Counters

69. Job Progress

70. All Tasks

71. Task Details

72. Task Counters

73. Task Logs

74. MapReduce Dataflow

75. MapReduce

76. Job Submission

77. Initialization

78. Scheduling

79. Execution

80. Map Task

81. Sort Buffer

82. Reduce Task

83. Next Generation
MapReduce
82

84. MapReduce Today
(Courtesy: Arun Murthy, Hortonworks)

85. Why ?
• Scalability Limitations today
• Maximum cluster size: 4000 nodes
• Maximum Concurrent tasks: 40,000
• Job Tracker SPOF
• Fixed map and reduce containers (slots)
• Punishes pleasantly parallel apps
84

86. Why ? (contd)
• MapReduce is not suitable for every
application
• Fine-Grained Iterative applications
• HaLoop: Hadoop in a Loop
• Message passing applications
• Graph Processing
85

87. Requirements
• Need scalable cluster resources manager
• Separate scheduling from resource
management
• Multi-Lingual Communication Protocols
86

88. Bottom Line
• @techmilind #mrng (MapReduce, Next
Gen) is in reality, #rmng (Resource Manager,
Next Gen)
• Expect different programming paradigms to
be implemented
• Including MPI (soon)
87

89. Architecture
(Courtesy: Arun Murthy, Hortonworks)

90. The New World
• Resource Manager
• Allocates resources (containers) to applications
• Node Manager
• Manages containers on nodes
• Application Master
• Specific to paradigm e.g. MapReduce application master,
MPI application master etc
89

91. Container
• In current terminology: A Task Slot
• Slice of the node s hardware resources
• #of cores, virtual memory, disk size, disk and
network bandwidth etc
• Currently, only memory usage is sliced
90