Big Data com Python

BigData with Python
A gentle and simple introduction
Marcel Caraciolo
@marcelcaraciolo
Developer, Cientist, contributor to the Crab recsys project,
works with Python for 6 years, interested at mobile,
education, machine learning and dataaaaa!
Recife, Brazil - http://aimotion.blogspot.com

https://class.coursera.org/datasci-001/
Disclaimer
Alguns slides foram
retirados das apts
do curso de Bill Howe -
Introduction to Data
Science

About me
Co-founder of Crab - Python recsys library
Cientist Chief at Atepassar, e-learning social network
Co-Founder and Instructor of PyCursos, teaching Python on-line
Co-Founder of Pingmind, on-line infrastructure for MOOC’s
Interested at Python, mobile, e-learning and machine learning!

Big Data
“Big Data is any data that is expensive to manage
and hard to extract value from.”
Michael Franklin Thomas M. Siebel
Professor of Computer Science Director of the Algorithms,
Machines and People Lab University of Berkeley

Big Data
Erik Larson, 1989, Harper’s magazine
“The keepers of big data say they do it for the
consumer’s benefit. But data have a way of
being used for purposes other than originally
intended.”

!"#$%"$#&'(')")(
*$+,(!"#$%"$#&'(')")(
!"#$%&'(%$)*%+%),-+.),#$$)
*#/(#*)01.#2%03)
!4'1-%$)*%+%5)6$'70)5)1$-18)0+9#%25)
:%1.-(#)7#(#9%+#*5);2$)#+1<3)
Big Data, muitos dados.

Challenges
Volume
the size of the data
Velocity
the latency of data processing relative to the
growing demand for interactivity
Variety
the diversity of sources, formats, quality,
structures

!"#$%&'&$%"()*+",*+$-./0$
.,12()$$
.&3"'4$ .)1,5"'4$
.&12)$
.&12)$

Where does big
data comes from ?
“data exhaust” from customers
new and pervasive sensors
the ability to “keep everything”

!"#$%&'(')*"+$#"''
!"#$%$&'(
)*+,"-.%(/%"01(
!"#$%&'"()%*+%
2"-&*3&(.4.+*(/"5607-8(
,"-./%('-0"&12%3/&124.(1%
500%6&'216%./7%!.281&%0#.(1%
91:.-/*1(9-.%'7/,(
;-(5*5"+'(/"5607-8(
9211/%7.&.%(1/&21%
:'#37%6&.&1%723;1%
<#.6)%!1-'2=%
,>!?@%
:'(3.#%AB!%
!"#$%*(.-.%'7/,(
<$8(=.&.(

Where does big
data comes from ?
photo in public domain

Where does big
data comes from ?
4/28/13 Bill Howe, UW 11

!"#$%&'()#*+$
$
,-.#(''/$
$
$011$*2))2'3$-.45#$67#&7$-7$'8$9-&."$:;<:=$
$<$23$>$?3@#&3#@$67#&7$"-5#$-$,-.#(''/$
$-..'63@$
$$
$9'&#$@"-3$>;$(2))2'3$A2#.#7$'8$.'3@#3@$BC#($
$)23/7=$3#C7$7@'&2#7=$()'D$A'7@7=$3'@#7=$A"'@'$
$-)(6*7=$#@.EF$7"-&#G$#-."$*'3@"E$
$
$H')G7$>;%I$'8$G-@-$8'&$-3-)J727=$-GG7$<:$!I$'8$
$.'*A&#77#G$G-@-$G-2)J$

!"#$%&'!
!
!())!*$++$,-!./&'/0!12)!*$++$,-!34$+5!6#&&6/!
!789!:$++$,-!/&4';<!=.&'$&/!4!345!
!>!"?!3464!@,'!4-4+5/$/!A&-&'46&3!34$+5!
!!
"<&!B',:+&*C!

!"#$%&''&(!
!
!)*+!#,-.'&/0!.'$('1!!2++!#,-.'&/0!3#$4'!
!5$6-44$/0!3#$4'!0&378!9-(!!*+++!.'$('!
!):!;<!0&#&!9-(!&/&78'3'!=$/$(&#$0!0&378!
!!
>(&03?-/&7!0&#&!'#-(&=$1!#$6,/3@.$'!A!&/&78'3'!
#--7'!B.'#!0-!/-#!C-(D!&#!#,$'$!'6&7$'!E!
!
>,$!F(-G7$4H!

What does
Scalable mean ?
Operationally:
In the past: Works even if data doesn’t fit in main memory
Now: Can make use of 1000s of cheap computers
Algorithmically:
In the past: If you have N data items, you must do no more
than Nm operations -- polynomial time algorithms
Now: If you have N data items, you must do no more than Nm/k operations,
for some large k -- “polynomial time algorithms must be parallelized”
Soon: If you have N data items, you should do no more than N * log(N) operations

Example
Given a set of DNA sequences, find all
sequences equal to GATTACGATATTA .

TACCTGCCGTAA = GATTACGATATTA?
GATTACGATATTA
time = 0
No.
Time = 0

Time = 1
CCCCCAATGAC = GATTACGATATTA?
GATTACGATATTA
time = 1
No.

Time = 17
GATTACGATATTA contains GATTACGATATTA?
GATTACGATATTA
time = 17
Yes!
Send it to the output.

GATTACGATATTA
40 records, 40 comparions
N records, N comparisons
The algorithmic complexity is order N: O(N)

GATTACGATATTA
AAAATCCTGCA
AAACGCCTGCA
TTTACGTCAA
TTTTCGTAATT
What if we sort the sequences?
What if we sort the sequences ?

CTGTACACAACCT < GATTACGATATTA
GATTACGATATTA
No match.
Skip to 75% mark
Start at the 50% mark
time = 0
0% 100%
CTGTACACAACCT
Time = 0

GGATACACATTTA > GATTACGATATTA
GATTACGATATTA
No match.
0% 100%
GGATACACATTTA

GATATTTTAAGC < GATTACGATATTA
GATTACGATATTA
No match.
Skip back to 68.75% mark
0% 100%
GATATTTTAAGC

GATTACGATATTA = GATTACGATATTA
GATTACGATATTA
0% 100%
Match!
Walk through the records until we
fail to match.

GATTACGATATTA
0% 100%
How many comparisons did we do?
40 records, only 4 comparisons
N records, log(N) comparisons
This algorithm is O(log(N)) Far better scalability

Relational
Databases
Relational Databases
•  Databases are good at Needle in Haystack problems:
–  Extracting small results from big datasets
–  Transparently provide old style scalability
–  Your query will always* finish, regardless of dataset size.
–  Indexes are easily built and automatically used when appropriate
CREATE INDEX seq_idx ON sequence(seq);
SELECT seq
FROM sequence
WHERE seq = GATTACGATATTA ;

Example
New task: Read Trimming
Given a set of DNA sequences,
trim the final n bps of each sequence
Generate a new dataset

TACCTGCCGTAA becomes TACCT
time = 0
Time = 0

Time = 1
CCCCCAATGAC becomes CCCCC
time = 1

Time = 17
GATTACGATATTA becomes GATTA
time = 17

Can we use an index?
No. We have to touch every record no matter what.
The task is fundamentally O(N)
Can we do any better?

Time = 7
time = 7
How much time did this take?
7 cycles
40 records, 6 workers
O(N/k)
time = 7
How much time did this ta
7 cycles
40 records, 6 workers
O(N/k)

Schematic of a Parallel “Read Trimming” Task
You are given short
“reads”: genomic
sequences about
35-75 characters each
Distribute the reads
among k computers
f f f f f f
f is a function to
trim a read; apply it
to every item
Now we have a big
distributed set of
trimmed reads
You are given short
sequences about
among k computers
f f f f f f
f is a function to
to every item
Now we have a big
distributed set of
trimmed reads
You are given short
sequences about
among k computers
f f f f f f
f is a function to
to every item
Now we have a big
distributed set of
trimmed reads
You are given short
sequences about
among k computers
f f f f f f
f is a function to
to every item
Now we have a big
distributed set of
trimmed reads

You are given TIFF
images
Distribute the images
among k computers
f f f f f f
f is a function to
convert TIFF to
PNG; apply it to
every item
Now we have a big
distributed set of
converted images
New Task: Convert 405k TIFF images to PNG
http://open.blogs.nytimes.com/2008/05/21/the-new-york-times-archives-amazon-web-services-timesmachine/
Convert 405k TIFF images to PNG

You have sets of
parameters for
thousands of small
simulations
Divide the parameter
sets among k
computers
f f f f f f
f is runs the
simulation and
produces some
output; apply it to
every item
Now we have a big
distributed set of
simulation results
New Task: Run thousands of simulations
http://escience.washington.edu/get-help-now/dave-williams-simulating-muscle-dynamics-cloud
Run thousands of simulations

You have millions
of documents
Distribute the
documents among k
computers
f f f f f f
f finds the most
common word in a
single document
Now we have a big
distributed list of
(doc_id, word) pairs
Find the most common word in each document
Find the most common word in each
document

Consider a slightly more general program
to compute the word frequency of every
word in a singe documentConsider a slightly more general program to compute the
word frequency of every word in a single document
Abridged Declaration of Independence
A Declaration By the Representatives of the United States of America, in General Congress Assembled.
When in the course of human events it becomes necessary for a people to advance from that subordination in
which they have hitherto remained, and to assume among powers of the earth the equal and independent station
to which the laws of nature and of nature's god entitle them, a decent respect to the opinions of mankind
requires that they should declare the causes which impel them to the change.
We hold these truths to be self-evident; that all men are created equal and independent; that from that equal
creation they derive rights inherent and inalienable, among which are the preservation of life, and liberty, and
the pursuit of happiness; that to secure these ends, governments are instituted among men, deriving their just
power from the consent of the governed; that whenever any form of government shall become destructive of
these ends, it is the right of the people to alter or to abolish it, and to institute new government, laying it's
foundation on such principles and organizing it's power in such form, as to them shall seem most likely to effect
their safety and happiness. Prudence indeed will dictate that governments long established should not be
changed for light and transient causes: and accordingly all experience hath shewn that mankind are more
disposed to suffer while evils are sufferable, than to right themselves by abolishing the forms to which they are
accustomed. But when a long train of abuses and usurpations, begun at a distinguished period, and pursuing
invariably the same object, evinces a design to reduce them to arbitrary power, it is their right, it is their duty, to
throw off such government and to provide new guards for future security. Such has been the patient sufferings
of the colonies; and such is now the necessity which constrains them to expunge their former systems of
government. the history of his present majesty is a history of unremitting injuries and usurpations, among which
no one fact stands single or solitary to contradict the uniform tenor of the rest, all of which have in direct object
the establishment of an absolute tyranny over these states. To prove this, let facts be submitted to a candid
world, for the truth of which we pledge a faith yet unsullied by falsehood.
(people, 2)
(government, 6)
(assume, 1)
(history, 2)
…

You have millions
of documents
Distribute the
documents among k
computers
f f f f f f
For each document
f returns a set of
(word, freq) pairs
Compute the word frequency of 5M documents
Now we have a big
distributed list of
sets of word freqs.

There’s a pattern...
There’s a pattern here….
•  A function that maps a read to a trimmed read
•  A function that maps a TIFF image to a PNG image
•  A function that maps a set of parameters to a simulation result
•  A function that maps a document to its most common word
•  A function that maps a document to a histogram of word
frequencies

What if we want to compute the word
frequency across all documents?
US Constitution Declaration of
Independence
Articles of
Confederation
(people, 78)
(government, 123)
(assume, 23)
(history, 38)
…

You have millions
of documents
Distribute the
documents among
k computers
map map map map map map
For each document,
return a set of (word,
freq) pairs
Compute the word frequency across 5M documents
Now what?
•  But we don’t want a bunch of little histograms –
we want one big histogram.
•  How can we make sure that a single computer
has access to every occurrence of a given word
regardless of which document it appeared in?
Compute the word frequency across 5M docs

Distribute the
documents among k
computers
map map map map map map
For each document,
return a set of
(word, freq) pairs
Now we have a big
distributed list of
sets of word freqs.
reduce reduce reduce reduce
Now just count the
occurrences of each word
44 3
We have our
distributed histogram
Compute the word frequency across 5M documents

5/15/13 Bill Howe, eScience Institute 11
Some distributed algorithm…
Map
(Shuffle)
Reduce

MapReduce Programming Model
•  Input & Output: each a set of key/value pairs
•  Programmer speciﬁes two functions:
–  Processes input key/value pair
–  Produces set of intermediate pairs

–  Combines all intermediate values for a particular key
–  Produces a set of merged output values (usually just one)
map (in_key, in_value) -> list(out_key, intermediate_value)
reduce (out_key, list(intermediate_value)) -> list(out_value)
Inspired by primitives from functional programming
languages such as Lisp, Scheme, and Haskell
slide source: Google, Inc.

Example: What does this do?
map(String input_key, String input_value):
// input_key: document name
// input_value: document contents
for each word w in input_value:
EmitIntermediate(w, 1);

reduce(String output_key, Iterator intermediate_values):
// output_key: word
// output_values: ????
int result = 0;
for each v in intermediate_values:
result += v;
Emit(result);
slide source: Google, Inc.

Example
Example: Document Processing

Word length histogram example
Example: Document Processing
Example: Word length histogram
How many big , medium , and small words are used?

A Declaration By the Representatives of the United States of America, in General
Congress Assembled.
When in the course of human events it becomes necessary for a people to advance from
that subordination in which they have hitherto remained, and to assume among powers of
the earth the equal and independent station to which the laws of nature and of nature's
god entitle them, a decent respect to the opinions of mankind requires that they should
declare the causes which impel them to the change.
We hold these truths to be self-evident; that all men are created equal and independent;
that from that equal creation they derive rights inherent and inalienable, among which are
the preservation of life, and liberty, and the pursuit of happiness; that to secure these
ends, governments are instituted among men, deriving their just power from the consent
of the governed; that whenever any form of government shall become destructive of these
ends, it is the right of the people to alter or to abolish it, and to institute new government,
laying it's foundation on such principles and organizing it's power in such form, as to
them shall seem most likely to effect their safety and happiness. Prudence indeed will
dictate that governments long established should not be changed for light and transient
causes: and accordingly all experience hath shewn that mankind are more disposed to
suffer while evils are sufferable, than to right themselves by abolishing the forms to
which they are accustomed. But when a long train of abuses and usurpations, begun at a
distinguished period, and pursuing invariably the same object, evinces a design to reduce
them to arbitrary power, it is their right, it is their duty, to throw off such government and
to provide new guards for future security. Such has been the patient sufferings of the
colonies; and such is now the necessity which constrains them to expunge their former
systems of government. the history of his present majesty is a history of unremitting
injuries and usurpations, among which no one fact stands single or solitary to contradict
the uniform tenor of the rest, all of which have in direct object the establishment of an
absolute tyranny over these states. To prove this, let facts be submitted to a candid world,
for the truth of which we pledge a faith yet unsullied by falsehood.
Big = Yellow = 10+ letters
Medium = Red = 5..9 letters
Small = Blue = 2..4 letters
Tiny = Pink = 1 letter

Congress Assembled.
Split the document into
chunks and process each
chunk on a different
computer
Chunk 1
Chunk 2

(yellow, 20)
(red, 71)
(blue, 93)
(pink, 6 )
Congress Assembled.
Map Task 1
(204 words)
Map Task 2
(190 words)
(key, value)
(yellow, 17)
(red, 77)
(blue, 107)
(pink, 3)

(yellow, 17)
(red, 77)
(blue, 107)
(pink, 3)
(yellow, 20)
(red, 71)
(blue, 93)
(pink, 6 )
Reduce tasks
(yellow, 17)
(yellow, 20)
(red, 77)
(red, 71)
(blue, 93)
(blue, 107)
(pink, 6)
(pink, 3)
Congress Assembled.
Map task 1
Map task 2
Shuffle step
(yellow, 37)
(red, 148)
(blue, 200)
(pink, 9)

MR Phases
!"#$%&'(")$*+&,&
MR Phases
•  Each Map and Reduce task has multiple phases:

MR Phases
MAP REDUCE
(w1,1)
(w2,1)
(w1,1)
…
(w1,1)
(w2,1)
…
(did1,v1)
(did2,v2)
(did3,v3)
. . . .
(w1, (1,1,1,…,1))
(w2, (1,1,…))
(w3,(1…))
…
…
…
…
(w1, 25)
(w2, 77)
(w3, 12)
…
…
…
…
Shuffle
!"#$%&'()%"**$"(+%,&-.$/%%
012%3',%45+,%+$3)%6&78%9:;%

MR Phases
Hadoop in
One Slide
src: Huy Vo,
NYU Poly

Taxonomy of
Parallel Architectures
Taxonomy of Parallel Architectures
Easiest to program,
but $$$$
Scales to 1000s of computers

Large-Scale Data Processing
•  Many tasks process big data, produce big data
•  Want to use hundreds or thousands of CPUs
–  ... but this needs to be easy
–  Parallel databases exist, but they are expensive, difficult to set
up, and do not necessarily scale to hundreds of nodes.
•  MapReduce is a lightweight framework, providing:
–  Automatic parallelization and distribution
–  Fault-tolerance
–  I/O scheduling
–  Status and monitoring

MapReduce Contemporaries
•  Dryad (Microsoft)
–  Relational Algebra
•  Pig (Yahoo)
–  Near Relational Algebra over MapReduce
•  HIVE (Facebook)
–  SQL over MapReduce
•  Cascading
–  Relational Algebra
•  Clustera
–  U of Wisconsin
•  Hbase
–  Indexing on HDFS
MapReduce Contemporaries

Talk is cheap,
Show me the code

Examples
More Examples: Build an Inverted Index
Input:
tweet1, (“I love pancakes for breakfast”)
tweet2, (“I dislike pancakes”)
tweet3, (“What should I eat for breakfast?”)
tweet4, (“I love to eat”)
Desired output:
“pancakes”, (tweet1, tweet2)
“breakfast”, (tweet1, tweet3)
“eat”, (tweet3, tweet4)
“love”, (tweet1, tweet4)
…

Examples
More Examples: Relational Join
Name SSN
Sue 999999999
Tony 777777777
EmpSSN DepName
999999999 Accounts
777777777 Sales
777777777 Marketing
Emplyee ⨝ Assigned Departments
Employee
Assigned Departments
Name SSN EmpSSN DepName
Sue 999999999 999999999 Accounts
Tony 777777777 777777777 Sales
Tony 777777777 777777777 Marketing

Examples
Relational Join in MapReduce: Before Map Phase
Name SSN
Sue 999999999
Tony 777777777
EmpSSN DepName
999999999 Accounts
777777777 Sales
777777777 Marketing
Employee
Assigned Departments
Key idea: Lump all the tuples
together into one dataset
Employee, Sue, 999999999
Employee, Tony, 777777777
Department, 999999999, Accounts
Department, 777777777, Sales
Department, 777777777, Marketing
What is this for?

Examples
Relational Join in MapReduce: Map Phase
Employee, Sue, 999999999
Employee, Tony, 777777777
Department, 999999999, Accounts
Department, 777777777, Sales
Department, 777777777, Marketing
key=999999999, value=(Employee, Sue, 999999999)
key=777777777, value=(Employee, Tony, 777777777)
key=999999999, value=(Department, 999999999, Accounts)
key=777777777, value=(Department, 777777777, Sales)
key=777777777, value=(Department, 777777777, Marketing)
why do we use this as the key?

Examples
Relational Join in MapReduce: Reduce Phase
key=999999999, values=[(Employee, Sue, 999999999),
(Department, 999999999, Accounts)]
key=777777777, values=[(Employee, Tony, 777777777),
(Department, 777777777, Sales),
(Department, 777777777, Marketing)]
Sue, 999999999, 999999999, Accounts
Tony, 777777777, 777777777, Sales
Tony, 777777777, 777777777, Marketing

Examples
Relational Join in MapReduce, again
5/15/13 Bill Howe, Data Science, Autumn 2012 19
Order(orderid, account, date)
1, aaa, d1
2, aaa, d2
3, bbb, d3
LineItem(orderid, itemid, qty)
1, 10, 1
1, 20, 3
2, 10, 5
2, 50, 100
3, 20, 1tagged with
relation name
Order
1, aaa, d1 ! 1 : “Order”, (1,aaa,d1)
2, aaa, d2 ! 2 : “Order”, (2,aaa,d2)
3, bbb, d3 ! 3 : “Order”, (3,bbb,d3)
Line
1, 10, 1 ! 1 : “Line”, (1, 10, 1)
1, 20, 3 ! 1 : “Line”, (1, 20, 3)
2, 10, 5 ! 2 : “Line”, (2, 10, 5)
2, 50, 100 ! 2 : “Line”, (2, 50, 100)
3, 20, 1 ! 3 : “Line”, (3, 20, 1)
Map
Reducer for key 1
“Order”, (1,aaa,d1)
“Line”, (1, 10, 1)
“Line”, (1, 20, 3)
(1, aaa, d1, 1, 10, 1)
(1, aaa, d1, 1, 20, 3)

ExamplesSimple Social Network Analysis:
Count Friends
Jim, Sue
Sue, Jim
Lin, Joe
Joe, Lin
Jim, Kai
Kai, Jim
Jim, Lin
Lin, Jim
Input Desired Output
Jim, 3
Lin, 2
Sue, 1
Kai, 1
Joe, 1
Jim, 1
Sue, 1
Lin, 1
Joe, 1
Jim, 1
Kai, 1
Jim, 1
Lin, 1
MAP REDUCE
Jim, (1, 1, 1)
Lin, (1, 1)
Sue, (1)
Kai, (1)
Joe, (1)
SHUFFLE

Examples
Matrix Multiplication
!" #" $" %&"
'" &" %#" !"
!" %&"
$" #"
%#" %&"
(" $"
X =
!" %)"
&#" $"

Examples
Matrix Multiply in MapReduce
C = A X B
A has dimensions L,M
B has dimensions M,N
•  In the map phase:
–  for each element (i,j) of A, emit ((i,k), A[i,j]) for k in 1..N
–  for each element (j,k) of B, emit ((i,k), B[j,k]) for i in 1..L
•  In the reduce phase, emit
–  key = (i,k)
–  value = Sumj (A[i,j] * B[j,k])
5/15/13 Bill Howe, Data Science, Autumn 2012 21

Examples
AB
X
A B
=
•  One reducer per output cell

Cluster Computing
•  Large number of commodity servers,
connected by high speed, commodity
network
•  Rack: holds a small number of servers
•  Data center: holds many racks
!"#$%&'(#))%'&*"&+!"
#$%&'()*+(",-(+./0"123145"67$%8('"1",-(+./09":;1;<"/0=>4""
/?"#@0@0A"/?"#$99@B("C$8$9(89;"D>"E$F$'$G$0"$0)"H==G$0""
),,(-.,(/0123,(456,7852(9,:3;-,(

Cluster Computing
•  Massive parallelism:
– 100s, or 1000s, or 10000s servers
– Many hours
•  Failure:
– If medium-time-between-failure is 1 year
– Then 10000 servers have one failure / hour
24slide src: Dan Suciu and Magda Balazinska

Distributed File System (DFS)
•  For very large files: TBs, PBs
•  Each file is partitioned into chunks,
typically 64MB
•  Each chunk is replicated several times
(!3), on different racks, for fault
tolerance
•  Implementations:
– Google’s DFS: GFS, proprietary
– Hadoop’s DFS: HDFS, open source
25slide src: Dan Suciu and Magda Balazinska

!"#$%&'%(#)**+%,%
!"#$%&"#'()*'(+(%"(&"#'(,-.%/#-/0,#'12,'
"(,3#'4-("#'*%4/,%&0/#*'-#$."'5,2-#44%)3'
2)'(')#/62,7'21'-2882*%/9'.(,*6(,#:'

!"#$%&'()"'*&+&*'",)-&$('
!"#$%%!&'((#)&#&*!+)(,-%.
.//'$)0(,123(),4'
5('%#4')0&')6'(%&'4(,)07&4('&$)'484(&94':1(%'*#,7&'0)')6'432'",)-&$(4'
;#%))'%#4')0&')6'(%&'2177&4('104(#**#<)0'=#>))"'
?300107'@///4')6'4&,+&,4')0'=#>))"'

!"#$%&'()*+),)
!"##$%&'"()'*'+,-'.&/01&'*'23$'4,5%&6'
'
'78193:&1:08&'5&93;/'"##$%&<='
'''''
' ')&,819'>;$3;&'
-&'./0&)01)2.(00$)$&03'4/)

!"#$%&"#"$'$()&*$+"$,&-../$0"#-$1.2$
3+456.%+&7$-&*&$&8&79"+"$
:#;*$<+8+84=$/&>#28"$"#&2%)$
?&%)+8#$7.4$&8&79"+"$
@#.A"/&%+B&7$&8&79"+"$
@#8.<#$C8&79"+"$
D204$D+"%.E#29$
F2&0-$&8-$%.</7+&8%#$<&8&4#<#8*$
G+-#.$&8-$+<&4#$&8&79"+"$
:2#8-$C8&79"+"$

!"#$%&'&$()*##+$,$-#./$-0&1$
2$34)5#.637$$
2$8)9':##;$$
2$<##/-'$$
2$=>?$$
2$@0&.'A$
2$B)&1CD4$$
2$E'F$G#H;$I04'&$$
2$G)"##J$$
2$IF0K'H$
2$B0.;'*$0.$$
$$
$

!"#$$%&'($)*)+',&-&
!"#$%& '#()%*+,%-!./0&
12+34#&
!56%& 758&
96:;&
<;;=%%(%:&
0>;;(&
/?+@%& A;B5%& C25::&

!"#$%&'(&
!""#
$%&'()*+&,#-&.&/&010#

!"#$%&'()&&
!""#
!"#$$%&#'()*'+,)-#&./-&(0()-1&

!"#$$%&'$(%$)*)+,&-&
.$/&+0"12*0&
3",2&30"12*0&
4"(*&4$#*&
5"%&6*#71*&
!89:&&&&

MapReduce
http://labs.google.com/papers/mapreduce.html

Exemplo de MapReduce
from mrjob.job import MRJob
import re
WORD_RE = re.compile(r"[w']+")
class MRWordFreqCount(MRJob):
def mapper(self, _, line):
for word in WORD_RE.findall(line):
yield (word.lower(), 1)
def reducer(self, word, counts):
yield (word, sum(counts))
if __name__ == '__main__':
MRWordFreqCount().run()

Projeto MrJob
Criado pela Equipe de Engenharia doYelp
Totalmente Open-Source
Todo em Python
Utiliza Map-Reduce para Processamento
Permite rodar tanto no Amazon EMR como no Hadoop

Objetivos do MrJobs
Se você quer aprender MapReduce, ele é para você
Se você tem um problema cavalar e precisa de muito
processamento e não está aﬁm de mexer em Hadoop
Se você já tem um cluster Hadoop e quer rodar scripts
Python
Se você quer migrar seu código Python do Hadoop
para o EMR
Se você não quer escrever Python (Impossível!), não é
para você!

Passos importantes
sudo easy_install mrjob

Vamos a uma demo...
Texto da posse de Obama em 2009.

Mais informações
http://packages.python.org/mrjob/
https://github.com/Yelp/mrjob

Distributed Computing with mrJob
https://github.com/Yelp/mrjob
Elsayed et al: Pairwise Document Similarity in Large Collections with MapReduce

Atepassar Recommendations
http://atepassar.com
Problema: Como recomendar novos amigos ?

marcel;jonas,maria,jose,amanda
maria;carol,fabiola,amanda,marcel
amanda;paula,patricia,maria,marcel
carol;maria,jose,patricia
fabiola;maria
paula;fabio,amanda
patricia;amanda,carol
jose;marcel,carol
jonas;marcel,fabio
fabio;jonas,paula
carla
"marcel" [["carol", 2], ["fabio", 1], ["fabiola", 1], ["patricia", 1], ["paula", 1]]
"maria" [["jose", 2], ["patricia", 2], ["jonas", 1], ["paula", 1]]
"patricia" [["maria", 2], ["jose", 1], ["marcel", 1], ["paula", 1]]
"paula" [["jonas", 1], ["marcel", 1], ["maria", 1], ["patricia", 1]]
"amanda" [["carol", 2], ["fabio", 1], ["fabiola", 1], ["jonas", 1], ["jose", 1]]
"carol" [["amanda", 2], ["marcel", 2], ["fabiola", 1]]
"fabio" [["amanda", 1], ["marcel", 1]]
"fabiola" [["amanda", 1], ["carol", 1], ["marcel", 1]]
"jonas" [["amanda", 1], ["jose", 1], ["maria", 1], ["paula", 1]]
"jose" [["maria", 2], ["amanda", 1], ["jonas", 1], ["patricia", 1]]
$python friends_recommender.py - r emr --num-ec2-instances 5 facebook_data.csv > output.dat

maria;carol,fabiola,amanda,marcel
amanda;paula,patricia,maria,marcel
fabiola;maria
paula;fabio,amanda
patricia;amanda,carol
jose;marcel,carol
jonas;marcel,fabio
fabio;jonas,paula
carla
"marcel" [["carol", 2], ["fabio", 1], ["fabiola", 1], ["patricia", 1], ["paula", 1]]
"maria" [["jose", 2], ["patricia", 2], ["jonas", 1], ["paula", 1]]
"patricia" [["maria", 2], ["jose", 1], ["marcel", 1], ["paula", 1]]
"paula" [["jonas", 1], ["marcel", 1], ["maria", 1], ["patricia", 1]]
"amanda" [["carol", 2], ["fabio", 1], ["fabiola", 1], ["jonas", 1], ["jose", 1]]
"carol" [["amanda", 2], ["marcel", 2], ["fabiola", 1]]
"fabio" [["amanda", 1], ["marcel", 1]]
"fabiola" [["amanda", 1], ["carol", 1], ["marcel", 1]]
"jonas" [["amanda", 1], ["jose", 1], ["maria", 1], ["paula", 1]]
"jose" [["maria", 2], ["amanda", 1], ["jonas", 1], ["patricia", 1]]
$python friends_recommender.py - r emr --num-ec2-instances 5 facebook_data.csv > output.dat
fabio;jonas,paula
...

fsim(u,i) = w1 f1(u,i) + w2 f2(u,i) + b,
https://gist.github.com/3970945#ﬁle_atepassar_recommender.py

Celery - para agendamento dos jobs coletores e
executores.
mrJob - para mapreduce e acesso ao Hadoop
MongoDb - para armazenamento das recomendações
Boto - acesso aos ﬁles do S3.

Projetos interessantes
Estrutura de dados para manipulação rápida
- slicing
- indexing
- subseting
Handling missing data
Agregações, Séries Temporais
Pandas: a data analysis library for Python,
poised to give R a run for its money… http://pandas.pydata.org/

Outro framework para computação distribuída com
Python com MapReduce.
Criado pelo Instituto Nokia.
Backend dele é escrito em Erlang (funcional,
concorrente e bem escalável!)
Não utiliza o FileSystem mais usado HDFS e sim um
novo padrão por eles (DDFS).
http://discoproject.org/

http://api.mongodb.org/python/2.0/examples/
map_reduce.html
Python & MongoDb
MongoDb - Banco de Dados Não relacional (NoSQL)
Possui suporte nativo built-in para fazer MapReduce.
Escrever o código em JS e não é muito legível e ﬁca preso ao
Mongo ...
>>> reduce = Code("function (key, values) {"
... " var total = 0;"
... " for (var i = 0; i < values.length; i++) {"
... " total += values[i];"
... " }"
... " return total;"
... "}")

https://github.com/klbostee/dumbo/wiki/Short-tutorial
Dumbo
Uma das primeiras bibliotecas em cima do
MapReduce e Python.
Complicado para começar e está desatualizada :(

Um wrapper em Python em cima do Hadoop
para computação distribuída.
http://pydoop.sourceforge.net/docs/index.html
Legal, mas dá um trabalho para conﬁgurar.

Mapreduce com Python na Google AppEngine
https://developers.google.com/appengine/docs/python/
dataprocessing/
Ainda experimental e ﬁca “preso” à plataforma
AppEngine.

Algoritmos de aprendizagem de máquina
Supervisionados & Não supervisionados
Pré-processamento, extração de dados
Avaliação de classiﬁcadores, Pipeline,
seleção de atributos.
http://scikit-learn.org/stable/

Processamento de linguagem natural
Várias ferramentas para tokenização,
pos tagging, named entity recognition,
classiﬁcadores, etc.
Vários corpus disponíveis!
http://nltk.org/

Processamento de linguagem natural
Várias ferramentas para tokenização,
pos tagging, named entity recognition,
classiﬁcadores, etc.
Vários corpus disponíveis!
http://nltk.org/
Fiquem de olho...
Pipeline for distributed Natural Language Processing, made in Python
https://github.com/NAMD/pypln

Our Project’s Home Page
http://muricoca.github.com/crab

Future Releases
Planned Release 0.13
New home for python-recsys:
https://github.com/python-recsys/crab
Planned Release 0.14
Support to Item-Based Recommenders using MapReduce with MrJob
New commiters: vinnigracindo, ocelma, fcurella

Join us!
1. Read our Wiki Page
https://github.com/muricoca/crab/wiki/Developer-Resources
2. Check out our current sprints and open issues
https://github.com/muricoca/crab/issues
3. Forks, Pull Requests mandatory
4. Join us at irc.freenode.net #muricoca or at our
discussion list
http://groups.google.com/group/scikit-crab

Vários outros ...
Matplotlib
StatsModelsScipy
Numpy
Matplotlib
NetworkX
PyBrain
milk
Orange

Livros recomendados
http://shop.oreilly.com/product/0636920022640.do?cmp=il-radar-ebooks-big-data-now-radar
For free...

Livros recomendados
For free...
http://infolab.stanford.edu/~ullman/mmds/book.pdf

Artigos recomendados
For free...
http://aimotion.blogspot.com.br/2012/10/atepassar-
recommendations-recommending.html
http://aimotion.blogspot.com.br/2012/08/introduction-to-
recommendations-with.html

Artigos recomendados
For free...
http://aimotion.blogspot.com.br/2012/10/atepassar-
recommendations-recommending.html
http://aimotion.blogspot.com.br/2012/08/introduction-to-
recommendations-with.html
https://github.com/marcelcaraciolo/recsys-mapreduce-mrjob

Big Data com Python

More Related Content

What's hot (20)

Viewers also liked (20)

Similar to Big Data com Python (20)

More from Marcel Caraciolo (20)

Recently uploaded (20)

Big Data com Python