Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases Kansas State University Olathe Tuesday, 12 August 2014 William H. Hsu Laboratory for Knowledge Discovery in Databases, Kansas State University Acknowledgements K-State Manhattan: Majed Alsadhan, Scott Finkeldei, Kyle Hudson, Surya Teja Kallumadi K-State Olathe: Dr. Prema Arasu, Dana Reinert, Paige Adams, Cathy Danahy, Angela Cummins, Emily Surdez, Quentin New, Amy Burgess Big Data Workshop: Day 1 Part II – Beginner Tutorial on MapReduce

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases What is MapReduce? A programming model (& its associated implementation) For processing large data set Exploits large set of commodity computers Executes process in distributed manner Offers high degree of transparencies In other words:  simple and maybe suitable for your tasks !!! © 2006, H. Setiawan, National University of Singapore

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases Distributed Grep Very big data Split data grep matches cat All matches

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases Distributed Word Count Very big data Split data count merge merged count

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases Map Reduce Map:  Accepts input key/value pair  Emits intermediate key/value pair Reduce :  Accepts intermediate key/value* pair  Emits output key/value pair Very big data Result MAPMAP REDUCEREDUCE Partitioning Function

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases Partitioning Function

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases Partitioning Function (2) Default : hash(key) mod R Guarantee:  Relatively well-balanced partitions  Ordering guarantee within partition Distributed Sort  Map: emit(key,value)  Reduce (with R=1): emit(key,value)

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases MapReduce Distributed Grep  Map: if match(value,pattern) emit(value,1)  Reduce: emit(key,sum(value*)) Distributed Word Count  Map: for all w in value do emit(w,1)  Reduce: emit(key,sum(value*))

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases MapReduce Transparencies Plus Google Distributed File System : Parallelization Fault-tolerance Locality optimization Load balancing

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases Suitable for your task if Have a cluster Working with large dataset Working with independent data (or assumed) Can be cast into map and reduce

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases MapReduce outside Google Hadoop (Java)  Emulates MapReduce and GFS The architecture of Hadoop MapReduce and DFS is master/slave MasterSlave MapReducejobtrackertasktracker DFSnamenodedatanode

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases Example Word Count (1) Map public static class MapClass extends MapReduceBase implements Mapper { private final static IntWritable one = new IntWritable(1); private Text word = new Text(); public void map(WritableComparable key, Writable value, OutputCollector output, Reporter reporter) throws IOException { String line = ((Text)value).toString(); StringTokenizer itr = new StringTokenizer(line); while (itr.hasMoreTokens()) { word.set(itr.nextToken()); output.collect(word, one); }

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases Example Word Count (2) Reduce public static class Reduce extends MapReduceBase implements Reducer { public void reduce(WritableComparable key, Iterator values, OutputCollector output, Reporter reporter) throws IOException { int sum = 0; while (values.hasNext()) { sum += ((IntWritable) values.next()).get(); } output.collect(key, new IntWritable(sum)); }

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases Example Word Count (3) Main public static void main(String[] args) throws IOException { //checking goes here JobConf conf = new JobConf(); conf.setOutputKeyClass(Text.class); conf.setOutputValueClass(IntWritable.class); conf.setMapperClass(MapClass.class); conf.setCombinerClass(Reduce.class); conf.setReducerClass(Reduce.class); conf.setInputPath(new Path(args[0])); conf.setOutputPath(new Path(args[1])); JobClient.runJob(conf); }

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases One time setup set hadoop-site.xml and slaves Initiate namenode Run Hadoop MapReduce and DFS Upload your data to DFS Run your process… Download your data from DFS

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases Summary A simple programming model for processing large dataset on large set of computer cluster Fun to use, focus on problem, and let the library deal with the messy detail

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases References Original paper ( On wikipedia ( Hadoop – MapReduce in Java ( Starfish - MapReduce in Ruby (

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases What is Cloud Computing? “Cloud” refers to large Internet services like Google, Yahoo, etc that run on 10,000’s of machines More recently, “cloud computing” refers to services by these companies that let external customers rent computing cycles on their clusters  Amazon EC2: virtual machines at 10¢/hour, billed hourly  Amazon S3: storage at 15¢/GB/month Attractive features:  Scale: up to 100’s of nodes  Fine-grained billing: pay only for what you use  Ease of use: sign up with credit card, get root access © 2009, M. Zaharia, University of California – Berkeley

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases What is MapReduce? Simple data-parallel programming model designed for scalability and fault-tolerance Pioneered by Google  Processes 20 petabytes of data per day Popularized by open-source Hadoop project  Used at Yahoo!, Facebook, Amazon, … © 2009, M. Zaharia, University of California – Berkeley

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases What is MapReduce used for? At Google:  Index construction for Google Search  Article clustering for Google News  Statistical machine translation At Yahoo!:  “Web map” powering Yahoo! Search  Spam detection for Yahoo! Mail At Facebook:  Data mining  Ad optimization  Spam detection

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases Example: Facebook Lexicon

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases Example: Facebook Lexicon

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases What is MapReduce used for? In research:  Astronomical image analysis (Washington)  Bioinformatics (Maryland)  Analyzing Wikipedia conflicts (PARC)  Natural language processing (CMU)  Particle physics (Nebraska)  Ocean climate simulation (Washington) 

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases Outline MapReduce architecture Example applications Getting started with Hadoop Higher-level languages over Hadoop: Pig and Hive Amazon Elastic MapReduce

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases MapReduce Design Goals 1.Scalability to large data volumes:  1000’s of machines, 10,000’s of disks 2.Cost-efficiency:  Commodity machines (cheap, but unreliable)  Commodity network  Automatic fault-tolerance (fewer administrators)  Easy to use (fewer programmers)

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases Typical Hadoop Cluster Aggregation switch Rack switch 40 nodes/rack, nodes in cluster 1 Gbps bandwidth within rack, 8 Gbps out of rack Node specs (Yahoo terasort): 8 x 2GHz cores, 8 GB RAM, 4 disks (= 4 TB?) Image from pdf

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases Typical Hadoop Cluster Image from pdf

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases Challenges 1.Cheap nodes fail, especially if you have many  Mean time between failures for 1 node = 3 years  Mean time between failures for 1000 nodes = 1 day  Solution: Build fault-tolerance into system 1.Commodity network = low bandwidth  Solution: Push computation to the data 1.Programming distributed systems is hard  Solution: Data-parallel programming model: users write “map” & “reduce” functions, system distributes work and handles faults

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases Hadoop Components Distributed file system (HDFS)  Single namespace for entire cluster  Replicates data 3x for fault-tolerance MapReduce framework  Executes user jobs specified as “map” and “reduce” functions  Manages work distribution & fault- tolerance

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases Hadoop Distributed File System Files split into 128MB blocks Blocks replicated across several datanodes (usually 3) Single namenode stores metadata (file names, block locations, etc) Optimized for large files, sequential reads Files are append-only Namenode Datanodes File1

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases MapReduce Programming Model Data type: key-value records Map function: (K in, V in )  list(K inter, V inter ) Reduce function: (K inter, list(V inter ))  list(K out, V out )

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases Example: Word Count def mapper(line): foreach word in line.split(): output(word, 1) def reducer(key, values): output(key, sum(values))

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases Word Count Execution the quick brown fox the fox ate the mouse how now brown cow Map Reduc e brown, 2 fox, 2 how, 1 now, 1 the, 3 ate, 1 cow, 1 mouse, 1 quick, 1 the, 1 brown, 1 fox, 1 quick, 1 the, 1 fox, 1 the, 1 how, 1 now, 1 brown, 1 ate, 1 mouse, 1 cow, 1 InputMapShuffle & SortReduceOutput

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases MapReduce Execution Details Single master controls job execution on multiple slaves Mappers preferentially placed on same node or same rack as their input block  Minimizes network usage Mappers save outputs to local disk before serving them to reducers  Allows recovery if a reducer crashes  Allows having more reducers than nodes

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases An Optimization: The Combiner def combiner(key, values): output(key, sum(values)) A combiner is a local aggregation function for repeated keys produced by same map Works for associative functions like sum, count, max Decreases size of intermediate data Example: map-side aggregation for Word Count:

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases Word Count with Combiner InputMap & CombineShuffle & SortReduceOutput the quick brown fox the fox ate the mouse how now brown cow Map Reduc e brown, 2 fox, 2 how, 1 now, 1 the, 3 ate, 1 cow, 1 mouse, 1 quick, 1 the, 1 brown, 1 fox, 1 quick, 1 the, 2 fox, 1 how, 1 now, 1 brown, 1 ate, 1 mouse, 1 cow, 1

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases Fault Tolerance in MapReduce 1. If a task crashes:  Retry on another node  OK for a map because it has no dependencies  OK for reduce because map outputs are on disk  If the same task fails repeatedly, fail the job or ignore that input block (user-controlled)  Note: For these fault tolerance features to work, your map and reduce tasks must be side-effect- free

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases Fault Tolerance in MapReduce 2. If a node crashes:  Re-launch its current tasks on other nodes  Re-run any maps the node previously ran  Necessary because their output files were lost along with the crashed node

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases Fault Tolerance in MapReduce 3. If a task is going slowly (straggler):  Launch second copy of task on another node (“speculative execution”)  Take the output of whichever copy finishes first, and kill the other  Surprisingly important in large clusters  Stragglers occur frequently due to failing hardware, software bugs, misconfiguration, etc  Single straggler may noticeably slow down a job

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases Takeaways By providing a data-parallel programming model, MapReduce can control job execution in useful ways:  Automatic division of job into tasks  Automatic placement of computation near data  Automatic load balancing  Recovery from failures & stragglers User focuses on application, not on complexities of distributed computing

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases Outline MapReduce architecture Example applications Getting started with Hadoop Higher-level languages over Hadoop: Pig and Hive Amazon Elastic MapReduce

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases 1. Search Input: (lineNumber, line) records Output: lines matching a given pattern Map: if(line matches pattern): output(line) Reduce: identify function  Alternative: no reducer (map-only job)

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases pig sheep yak zebra aardvark ant bee cow elephant 2. Sort Input: (key, value) records Output: same records, sorted by key Map: identity function Reduce: identify function Trick: Pick partitioning function h such that k 1 h(k 1 )<h(k 2 ) Map Reduce ant, bee zebra aardvark, elephant cow pig sheep, yak [A-M] [N-Z]

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases 3. Inverted Index Input: (filename, text) records Output: list of files containing each word Map: foreach word in text.split(): output(word, filename) Combine: uniquify filenames for each word Reduce: def reduce(word, filenames): output(word, sort(filenames))

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases Inverted Index Example to be or not to be afraid, (12th.txt) be, (12th.txt, hamlet.txt) greatness, (12th.txt) not, (12th.txt, hamlet.txt) of, (12th.txt) or, (hamlet.txt) to, (hamlet.txt) hamlet.tx t be not afraid of greatness 12th.txt to, hamlet.txt be, hamlet.txt or, hamlet.txt not, hamlet.txt be, 12th.txt not, 12th.txt afraid, 12th.txt of, 12th.txt greatness, 12th.txt

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases 4. Most Popular Words Input: (filename, text) records Output: top 100 words occurring in the most files Two-stage solution:  Job 1:  Create inverted index, giving (word, list(file)) records  Job 2:  Map each (word, list(file)) to (count, word)  Sort these records by count as in sort job Optimizations:  Map to (word, 1) instead of (word, file) in Job 1  Count files in job 1’s reducer rather than job 2’s mapper  Estimate count distribution in advance and drop rare words

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases 5. Numerical Integration Input: (start, end) records for sub-ranges to integrate  Easy using custom InputFormat Output: integral of f(x) dx over entire range Map: def map(start, end): sum = 0 for(x = start; x < end; x += step): sum += f(x) * step output(“”, sum) Reduce: def reduce(key, values): output(key, sum(values))

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases Outline MapReduce architecture Example applications Getting started with Hadoop Higher-level languages over Hadoop: Pig and Hive Amazon Elastic MapReduce

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases Getting Started with Hadoop Download from hadoop.apache.orghadoop.apache.org To install locally, unzip and set JAVA_HOME Details: hadoop.apache.org/core/docs/current/quickstart.html hadoop.apache.org/core/docs/current/quickstart.html Three ways to write jobs:  Java API  Hadoop Streaming (for Python, Perl, etc)  Pipes API (C++)

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases Word Count in Java public class MapClass extends MapReduceBase implements Mapper { private final static IntWritable ONE = new IntWritable(1); public void map(LongWritable key, Text value, OutputCollector out, Reporter reporter) throws IOException { String line = value.toString(); StringTokenizer itr = new StringTokenizer(line); while (itr.hasMoreTokens()) { out.collect(new text(itr.nextToken()), ONE); }

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases Word Count in Java public class ReduceClass extends MapReduceBase implements Reducer { public void reduce(Text key, Iterator values, OutputCollector out, Reporter reporter) throws IOException { int sum = 0; while (values.hasNext()) { sum += values.next().get(); } out.collect(key, new IntWritable(sum)); }

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases Word Count in Java public static void main(String[] args) throws Exception { JobConf conf = new JobConf(WordCount.class); conf.setJobName("wordcount"); conf.setMapperClass(MapClass.class); conf.setCombinerClass(ReduceClass.class); conf.setReducerClass(ReduceClass.class); FileInputFormat.setInputPaths(conf, args[0]); FileOutputFormat.setOutputPath(conf, new Path(args[1])); conf.setOutputKeyClass(Text.class); // out keys are words (strings) conf.setOutputValueClass(IntWritable.class); // values are counts JobClient.runJob(conf); }

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases Word Count in Python with Hadoop Streaming import sys for line in sys.stdin: for word in line.split(): print(word.lower() + "\t" + 1) import sys counts = {} for line in sys.stdin: word, count = line.split("\t”) dict[word] = dict.get(word, 0) + int(count) for word, count in counts: print(word.lower() + "\t" + 1) Mapper.py: Reducer.py:

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases Outline MapReduce architecture Example applications Getting started with Hadoop Higher-level languages over Hadoop: Pig and Hive Amazon Elastic MapReduce

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases Motivation Many parallel algorithms can be expressed by a series of MapReduce jobs But MapReduce is fairly low-level: must think about keys, values, partitioning, etc Can we capture common “job building blocks”?

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases Pig Started at Yahoo! Research Runs about 30% of Yahoo!’s jobs Features:  Expresses sequences of MapReduce jobs  Data model: nested “bags” of items  Provides relational (SQL) operators (JOIN, GROUP BY, etc)  Easy to plug in Java functions  Pig Pen development environment for Eclipse

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases An Example Problem Suppose you have user data in one file, page view data in another, and you need to find the top 5 most visited pages by users aged Load Users Load Pages Filter by age Join on name Group on url Count clicks Order by clicks Take top 5 Example from

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases In MapReduce Example from

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases Users = load ‘users’ as (name, age); Filtered = filter Users by age >= 18 and age <= 25; Pages = load ‘pages’ as (user, url); Joined = join Filtered by name, Pages by user; Grouped = group Joined by url; Summed = foreach Grouped generate group, count(Joined) as clicks; Sorted = order Summed by clicks desc; Top5 = limit Sorted 5; store Top5 into ‘top5sites’; Example from In Pig Latin

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases Ease of Translation Notice how naturally the components of the job translate into Pig Latin. Load Users Load Pages Filter by age Join on name Group on url Count clicks Order by clicks Take top 5 Users = load … Filtered = filter … Pages = load … Joined = join … Grouped = group … Summed = … count()… Sorted = order … Top5 = limit … Example from

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases Ease of Translation Notice how naturally the components of the job translate into Pig Latin. Load Users Load Pages Filter by age Join on name Group on url Count clicks Order by clicks Take top 5 Users = load … Filtered = filter … Pages = load … Joined = join … Grouped = group … Summed = … count()… Sorted = order … Top5 = limit … Job 1 Job 2 Job 3 Example from

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases Hive Developed at Facebook Used for majority of Facebook jobs “Relational database” built on Hadoop  Maintains list of table schemas  SQL-like query language (HQL)  Can call Hadoop Streaming scripts from HQL  Supports table partitioning, clustering, complex data types, some optimizations

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases Sample Hive Queries SELECT p.url, COUNT(1) as clicks FROM users u JOIN page_views p ON (u.name = p.user) WHERE u.age >= 18 AND u.age <= 25 GROUP BY p.url ORDER BY clicks LIMIT 5; Find top 5 pages visited by users aged 18-25: Filter page views through Python script: SELECT TRANSFORM(p.user, p.date) USING 'map_script.py' AS dt, uid CLUSTER BY dt FROM page_views p;

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases Outline MapReduce architecture Example applications Getting started with Hadoop Higher-level languages over Hadoop: Pig and Hive Amazon Elastic MapReduce

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases Amazon Elastic MapReduce Provides a web-based interface and command-line tools for running Hadoop jobs on Amazon EC2 Data stored in Amazon S3 Monitors job and shuts down machines after use Small extra charge on top of EC2 pricing If you want more control over how you Hadoop runs, you can launch a Hadoop cluster on EC2 manually using the scripts in src/contrib/ec2

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases Elastic MapReduce Workflow

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases Elastic MapReduce Workflow

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases Elastic MapReduce Workflow

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases Elastic MapReduce Workflow

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases Conclusions MapReduce programming model hides the complexity of work distribution and fault tolerance Principal design philosophies:  Make it scalable, so you can throw hardware at problems  Make it cheap, lowering hardware, programming and admin costs MapReduce is not suitable for all problems, but when it works, it may save you quite a bit of time Cloud computing makes it straightforward to start using Hadoop (or other parallel software) at scale

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases MapReduce: Recap Programmers must specify: map (k, v) → * reduce (k’, v’) → *  All values with the same key are reduced together Optionally, also: partition (k’, number of partitions) → partition for k’  Often a simple hash of the key, e.g., hash(k’) mod n  Divides up key space for parallel reduce operations combine (k’, v’) → *  Mini-reducers that run in memory after the map phase  Used as an optimization to reduce network traffic The execution framework handles everything else… Adapted from slides © 2012, J. Lin & R. Jin

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases “Everything Else” The execution framework handles everything else…  Scheduling: assigns workers to map and reduce tasks  “Data distribution”: moves processes to data  Synchronization: gathers, sorts, and shuffles intermediate data  Errors and faults: detects worker failures and restarts Limited control over data and execution flow  All algorithms must expressed in m, r, c, p You don’t know:  Where mappers and reducers run  When a mapper or reducer begins or finishes  Which input a particular mapper is processing  Which intermediate key a particular reducer is processing

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases combine ba12c9ac52bc78 partition map k1k1 k2k2 k3k3 k4k4 k5k5 k6k6 v1v1 v2v2 v3v3 v4v4 v5v5 v6v6 ba12cc36ac52bc78 Shuffle and Sort: aggregate values by keys reduc e a15b27c298 r1r1 s1s1 r2r2 s2s2 r3r3 s3s3

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases Tools for Synchronization Cleverly-constructed data structures –Bring partial results together Sort order of intermediate keys –Control order in which reducers process keys Partitioner –Control which reducer processes which keys Preserving state in mappers and reducers –Capture dependencies across multiple keys and values

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases Preserving State Mapper object configure map close state one object per task Reducer object configure reduce close state one call per input key-value pair one call per intermediate key API initialization hook API cleanup hook

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases Scalable Hadoop Algorithms: Themes Avoid object creation  Inherently costly operation  Garbage collection Avoid buffering  Limited heap size  Works for small datasets, but won’t scale!

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases Importance of Local Aggregation Ideal scaling characteristics:  Twice the data, twice the running time  Twice the resources, half the running time Why can’t we achieve this?  Synchronization requires communication  Communication kills performance Thus… avoid communication!  Reduce intermediate data via local aggregation  Combiners can help

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases Shuffle and Sort Mapper Reducer other mappers other reducers circular buffer (in memory) spills (on disk) merged spills (on disk) intermediate files (on disk) Combiner

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases Word Count: Baseline What’s the impact of combiners?

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases Word Count: Version 1 Are combiners still needed?

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases Word Count: Version 2 Are combiners still needed? Key: preserve state across input key-value pairs!

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases Design Pattern for Local Aggregation “In-mapper combining”  Fold the functionality of the combiner into the mapper by preserving state across multiple map calls Advantages  Speed  Why is this faster than actual combiners? Disadvantages  Explicit memory management required  Potential for order-dependent bugs

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases Combiner Design Combiners and reducers share same method signature  Sometimes, reducers can serve as combiners  Often, not… Remember: combiner are optional optimizations  Should not affect algorithm correctness  May be run 0, 1, or multiple times Example: find average of all integers associated with the same key

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases Computing the Mean: Version 1 Why can’t we use reducer as combiner?

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases Computing the Mean: Version 2 Why doesn’t this work?

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases Computing the Mean: Version 3 Fixed?

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases Computing the Mean: Version 4 Are combiners still needed?

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases Algorithm Design: Running Example Term co-occurrence matrix for a text collection  M = N x N matrix (N = vocabulary size)  M ij : number of times i and j co-occur in some context (for concreteness, let’s say context = sentence) Why?  Distributional profiles as a way of measuring semantic distance  Semantic distance useful for many language processing tasks

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases MapReduce: Large Counting Problems Term co-occurrence matrix for a text collection = specific instance of a large counting problem –A large event space (number of terms) –A large number of observations (the collection itself) –Goal: keep track of interesting statistics about the events Basic approach –Mappers generate partial counts –Reducers aggregate partial counts How do we aggregate partial counts efficiently?

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases First Try: “Pairs” Each mapper takes a sentence:  Generate all co-occurring term pairs  For all pairs, emit (a, b) → count Reducers sum up counts associated with these pairs Use combiners!

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases Pairs: Pseudo-Code

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases “Pairs” Analysis Advantages  Easy to implement, easy to understand Disadvantages  Lots of pairs to sort and shuffle around (upper bound?)  Not many opportunities for combiners to work

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases Another Try: “Stripes” Idea: group together pairs into an associative array Each mapper takes a sentence:  Generate all co-occurring term pairs  For each term, emit a → { b: count b, c: count c, d: count d … } Reducers perform element-wise sum of associative arrays (a, b) → 1 (a, c) → 2 (a, d) → 5 (a, e) → 3 (a, f) → 2 a → { b: 1, c: 2, d: 5, e: 3, f: 2 } a → { b: 1, d: 5, e: 3 } a → { b: 1, c: 2, d: 2, f: 2 } a → { b: 2, c: 2, d: 7, e: 3, f: 2 } + Key: cleverly-constructed data structure brings together partial results

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases Stripes: Pseudo-Code

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases “Stripes” Analysis Advantages  Far less sorting and shuffling of key-value pairs  Can make better use of combiners Disadvantages  More difficult to implement  Underlying object more heavyweight  Fundamental limitation in terms of size of event space

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases Cluster size: 38 cores Data Source: Associated Press Worldstream (APW) of the English Gigaword Corpus (v3), which contains 2.27 million documents (1.8 GB compressed, 5.7 GB uncompressed)

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases Relative Frequencies How do we estimate relative frequencies from counts? Why do we want to do this? How do we do this with MapReduce?

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases f(B|A): “Stripes” Easy!  One pass to compute (a, *)  Another pass to directly compute f(B|A) a → {b 1 :3, b 2 :12, b 3 :7, b 4 :1, … }

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases f(B|A): “Pairs” For this to work:  Must emit extra (a, *) for every b n in mapper  Must make sure all a’s get sent to same reducer (use partitioner)  Must make sure (a, *) comes first (define sort order)  Must hold state in reducer across different key-value pairs (a, b 1 ) → 3 (a, b 2 ) → 12 (a, b 3 ) → 7 (a, b 4 ) → 1 … (a, *) → 32 (a, b 1 ) → 3 / 32 (a, b 2 ) → 12 / 32 (a, b 3 ) → 7 / 32 (a, b 4 ) → 1 / 32 … Reducer holds this value in memory

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases “Order Inversion” Common design pattern –Computing relative frequencies requires marginal counts –But marginal cannot be computed until you see all counts –Buffering is a bad idea! –Trick: getting the marginal counts to arrive at the reducer before the joint counts Optimizations –Apply in-memory combining pattern to accumulate marginal counts –Should we apply combiners?

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases Synchronization: Pairs vs. Stripes Approach 1: turn synchronization into an ordering problem  Sort keys into correct order of computation  Partition key space so that each reducer gets the appropriate set of partial results  Hold state in reducer across multiple key-value pairs to perform computation  Illustrated by the “pairs” approach Approach 2: construct data structures that bring partial results together  Each reducer receives all the data it needs to complete the computation  Illustrated by the “stripes” approach

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases Secondary Sorting MapReduce sorts input to reducers by key  Values may be arbitrarily ordered What if want to sort value also?  E.g., k → (v 1, r), (v 3, r), (v 4, r), (v 8, r)…

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases Secondary Sorting: Solutions Solution 1:  Buffer values in memory, then sort  Why is this a bad idea? Solution 2:  “Value-to-key conversion” design pattern: form composite intermediate key, (k, v 1 )  Let execution framework do the sorting  Preserve state across multiple key-value pairs to handle processing  Anything else we need to do?

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases Recap: Tools for Synchronization Cleverly-constructed data structures –Bring data together Sort order of intermediate keys –Control order in which reducers process keys Partitioner –Control which reducer processes which keys Preserving state in mappers and reducers –Capture dependencies across multiple keys and values

Computing & Information Sciences Kansas State University Kansas State University Olathe Workshop on Big Data – August, 2014 KSU Laboratory for Knowledge Discovery in Databases Issues and Tradeoffs Number of key-value pairs –Object creation overhead –Time for sorting and shuffling pairs across the network Size of each key-value pair –De/serialization overhead Local aggregation –Opportunities to perform local aggregation varies –Combiners make a big difference –Combiners vs. in-mapper combining –RAM vs. disk vs. network