Map Reduce Architecture

Adapted from Lectures by

Anand Rajaraman (Stanford Univ.)
and Dan Weld (Univ. of Washington)

Prasad L06MapReduce 1
Single-node architecture

CPU
Machine Learning, Statistics

Memory

“Classical” Data Mining

Disk

Prasad L06MapReduce 2
Commodity Clusters
 Web data sets can be very large
 Tens to hundreds of terabytes
 Cannot mine on a single server (why?)
 Standard architecture emerging:
 Cluster of commodity Linux nodes
 Gigabit ethernet interconnect
 How to organize computations on this
architecture?
 Mask issues such as hardware failure
 General Purpose Programming Model
Prasad L06MapReduce 3
Cluster Architecture
2-10 Gbps backbone between racks
1 Gbps between Switch
any pair of nodes
in a rack
Switch Switch

CPU CPU CPU CPU

Mem … Mem Mem … Mem

Disk Disk Disk Disk

Each rack contains 16-64 nodes

Prasad L06MapReduce 4
Stable storage
 First order problem: if nodes can fail,
how can we store data persistently?
 Answer: Distributed File System
 Provides global file namespace
 Google GFS; Hadoop HDFS; Kosmix KFS
 File replication (e.g., 3 copies)
 Typical usage pattern
 Huge files (100s of GB to TB)
 Data is rarely updated in place
 Reads and appends are common
Prasad L06MapReduce 5
Distributed File System
 Chunk Servers
 File is split into contiguous chunks
 Typically each chunk is 16-64MB
 Each chunk replicated (usually 2x or 3x)
 Try to keep replicas in different racks
 Master node
 a.k.a. Name Nodes in HDFS
 Stores metadata
 Might be replicated
 Client library for file access
 Talks to master to find chunk servers
 Connects directly to chunkservers to access data
Prasad L06MapReduce 6
 Motivation for MapReduce (why)
 Large-Scale Data Processing
 Want to use 1000s of CPUs
 But don’t want hassle of managing things

 MapReduce Architecture provides

 Automatic parallelization & distribution
 Fault tolerance
 I/O scheduling
 Monitoring & status updates
 Redundant computations (e.g., 3 processes)
Prasad L06MapReduce 7
 What is Map/Reduce

 Map/Reduce
 Programming model from LISP
 (and other functional languages)

 Many problems can be phrased this way

 Easy to distribute across nodes

 Nice retry/failure semantics
Prasad L06MapReduce 8
Map in LISP (Scheme)
 (map f list [list2 list3 …])
a tor
p e r
y o
a r
Un

 (map square ‘(1 2 3 4))

 (1 4 9 16)

Prasad L06MapReduce 9
Reduce in LISP (Scheme)
 (reduce f id list) ra tor
op e
a ry
Bin
 (reduce + 0 ‘(1 4 9 16))
 (+ 16 (+ 9 (+ 4 (+ 1 0)) ) )
 30

 (reduce + 0
(map square (map – l1 l2))))
Prasad L06MapReduce 10
Warm up: Word Count
 We have a large file of words, one
word to a line
 Count the number of times each
distinct word appears in the file

 Sample application: analyze web

server logs to find popular URLs

Prasad L06MapReduce 11
Word Count (2)
 Case 1: Entire file fits in memory
 Case 2: File too large for mem, but all
<word, count> pairs fit in mem
 Case 3: File on disk, too many distinct
words to fit in memory
 sort datafile | uniq –c

Prasad L06MapReduce 12
Word Count (3)
 To make it slightly harder, suppose
we have a large corpus of documents
 Count the number of times each
distinct word occurs in the corpus
words(docs/*) | sort | uniq -c
where words takes a file and outputs the
words in it, one to a line
 The above captures the essence of
MapReduce
 Great thing is it is naturally parallelizable
Prasad L06MapReduce 13
MapReduce
 Input: a set of key/value pairs
 User supplies two functions:
 map(k,v)  list(k1,v1)
 reduce(k1, list(v1))  v2
 (k1,v1) is an intermediate key/value
pair
 Output is the set of (k1,v2) pairs

Prasad L06MapReduce 14
Word Count using MapReduce
map(key, value):
// key: document name; value: text of document
for each word w in value:
emit(w, 1)

reduce(key, values):
// key: a word; values: an iterator over counts
result = 0
for each count v in values:
result += v
emit(key,result)
Prasad L06MapReduce 15
Count, map(key=url, val=contents):
Illustrated For each word w in contents, emit (w, “1”)
reduce(key=word, values=uniq_counts):
Sum all “1”s in values list
Emit result “(word, sum)”

see bob run see 1 bob 1

see spot bob 1 run 1
throw run 1 see 2
see 1 spot 1
spot 1 throw 1
throw 1

Prasad L06MapReduce 16
Model is Widely Applicable
MapReduce Programs In Google Source Tree

Example uses:
distributed grep distributed sort web link-graph reversal
web access log inverted index
term-vector / host
stats construction
document statistical machine
Prasad machine learning
L06MapReduce 17
clustering translation
 Implementation
Typical cluster: Overview

• 100s/1000s of 2-CPU x86 machines, 2-4 GB of

memory
• Limited bisection bandwidth
• Storage is on local IDE disks
• GFS: distributed file system manages data
(SOSP'03)
• Job scheduling system: jobs made up of tasks,
scheduler assigns tasks to machines

Implementation is a C++ library linked into

user programs

Prasad L06MapReduce 18
 Distributed Execution Overview
User
Program

fork fork fork

assign Master
assign
map reduce
Input Data Worker
write Output
local Worker File 0
Split 0 read
write
Split 1 Worker
Split 2 Output
Worker File 1
Worker remote
read,
sort
Prasad L06MapReduce 19
Data flow
 Input, final output are stored on a
distributed file system
 Scheduler tries to schedule map tasks
“close” to physical storage location of
input data
 Intermediate results are stored on
local FS of map and reduce workers
 Output is often input to another map
reduce task

Prasad L06MapReduce 20
Coordination
 Master data structures
 Task status: (idle, in-progress, completed)
 Idle tasks get scheduled as workers
become available
 When a map task completes, it sends the
master the location and sizes of its R
intermediate files, one for each reducer
 Master pushes this info to reducers
 Master pings workers periodically to
detect failures
Prasad L06MapReduce 21
Failures
 Map worker failure
 Map tasks completed or in-progress at
worker are reset to idle
 Reduce workers are notified when task is
rescheduled on another worker
 Reduce worker failure
 Only in-progress tasks are reset to idle
 Master failure
 MapReduce task is aborted and client is
notified
Prasad L06MapReduce 22
Execution

Prasad L06MapReduce 23
Parallel Execution

Prasad L06MapReduce 24
How many Map and Reduce jobs?
 M map tasks, R reduce tasks
 Rule of thumb:
 Make M and R much larger than the
number of nodes in cluster
 One DFS chunk per map is common
 Improves dynamic load balancing and
speeds recovery from worker failure
 Usually R is smaller than M, because
output is spread across R files

Prasad L06MapReduce 25
Combiners
 Often a map task will produce many
pairs of the form (k,v1), (k,v2), … for
the same key k
 E.g., popular words in Word Count
 Can save network time by pre-
aggregating at mapper
 combine(k1, list(v1))  v2
 Usually same as reduce function
 Works only if reduce function is
commutative and associative
Prasad L06MapReduce 26
Partition Function
 Inputs to map tasks are created by
contiguous splits of input file
 For reduce, we need to ensure that
records with the same intermediate
key end up at the same worker
 System uses a default partition
function e.g., hash(key) mod R
 Sometimes useful to override
 E.g., hash(hostname(URL)) mod R
ensures URLs from a host end up in the
same output file
Prasad L06MapReduce 27
Execution Summary
 How is this distributed?
1. Partition input key/value pairs into
chunks, run map() tasks in parallel
2. After all map()s are complete,
consolidate all emitted values for each
unique emitted key
3. Now partition space of output map keys,
and run reduce() in parallel
 If map() or reduce() fails, reexecute!

Prasad L06MapReduce 28
Logical flow of a MapReduce
programming model.

Prasad L06MapReduce 29
Prasad L06MapReduce 30
Exercise 1: Host size
 Suppose we have a large web corpus
 Let’s look at the metadata file
 Lines of the form (URL, size, date, …)
 For each host, find the total number
of bytes
 i.e., the sum of the page sizes for all
URLs from that host

Prasad L06MapReduce 31
Exercise 2: Distributed Grep
 Find all occurrences of the given
pattern in a very large set of files

Prasad L06MapReduce 32
Exercise 3: Graph reversal
 Given a directed graph as an
adjacency list:
src1: dest11, dest12, …
src2: dest21, dest22, …

 Construct the graph in which all the

links are reversed

Prasad L06MapReduce 34
Reverse Web-Link Graph
 Map
 For each URL linking to target, …
 Output <target, source> pairs
 Reduce
 Concatenate list of all source URLs
 Outputs: <target, list (source)> pairs

Prasad L06MapReduce 35
Hadoop
 An open-source implementation of
Map Reduce in Java
 Uses HDFS for stable storage
 Download from:
http://lucene.apache.org/hadoop/

Prasad L06MapReduce 37
Reading
 Jeffrey Dean and Sanjay Ghemawat,
MapReduce: Simplified Data Processing
on Large Clusters
http://labs.google.com/papers/mapreduce.html

 Sanjay Ghemawat, Howard Gobioff, and Shun-

Tak Leung, The Google File System
http://labs.google.com/papers/gfs.html

Prasad L06MapReduce 38
Conclusions

 MapReduce proven to be useful abstraction

 Greatly simplifies large-scale computations

 Fun to use:
 focus on problem,
 let library deal w/ messy details

Prasad L06MapReduce 39