Map-Reduce and the New Software Stack Advanced Search Techniques for Large Scale Data Analytics Pavel Zezula and Jan Sedmidubsky Masaryk University http://disa.fi.muni.cz MapReduce ■ Much of the course will be devoted to large scale computing for data mining ■ Challenges: ■ How to distribute computation? ■ Distributed/parallel programming is hard ■ Map-reduce addresses all of the above ■ Google's computational/data manipulation model ■ Elegant way to work with big data Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 2 Single Node Architecture Machine Learning, Statistics Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 3 Single Node Architecture Memory Classical" Data Mining Disk Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 3 Single Node Architecture CPU Memory Disk "Classical" Data Mining Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 3 Motivation: Google Example ■ 20+ billion web pages x 20KB = 400+ TB ■ 1 computer reads 30-35 MB/sec from disk ■ ~4 months to read the web ■ ~1,000 hard drives to store the web ■ Takes even more to do something useful with the data! ■ Today, a standard architecture for such problems is emerging: ■ Cluster of commodity Linux nodes ■ Commodity network (ethernet) to connect them Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 4 Cluster Architecture l Gbps between any pair of nodes in a rack Switch CPU Mem Disk Each rack contains 16-64 nodes In 2011 it was guestimated that Google had 1M machines, http://bit.lv/Shh0RO Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 5 Cluster Architecture 2-10 Gbps backbone between racks l Gbps between any pair of nodes in a rack CPU Mem Disk CPU Mem Disk Each rack contains 16-64 nodes In 2011 it was guestimated that Google had 1M machines, http://bit.lv/ShhORO Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 5 Large-scale Computing ■ Large-scale computing for data mining problems on commodity hardware ■ Challenges: ■ How do you distribute computation? ■ How can we make it easy to write distributed programs? ■ Machines fail: ■ One server may stay up 3 years (1,000 days) ■ If you have 1,000 servers, expect to loose 1/day ■ People estimated Google had ~1M machines in 2011 ■ 1,000 machines fail every day! Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 6 Idea and Solution ■ Issue: Copying data over a network takes time ■ Idea: ■ Bring computation close to the data ■ Store files multiple times for reliability Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 7 Idea and Solution ■ Issue: Copying data over a network takes time ■ Idea: ■ Bring computation close to the data ■ Store files multiple times for reliability ■ Map-reduce addresses these problems ■ Google's computational/data manipulation model ■ Elegant way to work with big data Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 7 Idea and Solution ■ Issue: Copying data over a network takes time ■ Idea: ■ Bring computation close to the data ■ Store files multiple times for reliability ■ Map-reduce addresses these problems ■ Google's computational/data manipulation model ■ Elegant way to work with big data ■ Storage Infrastructure - File system ■ Google: GFS. Hadoop: HDFS ■ Programming model ■ Map-Reduce Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 7 Storage Infrastructure ■ Problem: ■ If nodes fail, how to store data persistently? Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 8 Storage Infrastructure ■ Problem: ■ If nodes fail, how to store data persistently? ■ Answer: ■ Distributed File System: ■ Provides global file namespace ■ Google GFS; Hadoop HDFS; Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 8 Storage Infrastructure ■ Problem: ■ If nodes fail, how to store data persistently? ■ Answer: ■ Distributed File System: ■ Provides global file namespace ■ Google GFS; Hadoop HDFS; ■ Typical usage pattern ■ Huge files (100s of GB to TB) ■ Data is rarely updated in place ■ Reads and appends are common Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 8 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 Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 9 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 Node in Hadoop's HDFS ■ Stores metadata about where files are stored ■ Might be replicated Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 9 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 Node in Hadoop's HDFS ■ Stores metadata about where files are stored ■ Might be replicated ■ Client library for file access ■ Talks to master to find chunk servers ■ Connects directly to chunk servers to access data Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 9 Distributed File System ■ Reliable distributed file system ■ Data kept in "chunks" spread across machines ■ Each chunk replicated on different machines ■ Seamless recovery from disk or machine failure '0 c5 c2 c5 D1 '0 Chunk server 1 Chunk server 2 Chunk server 3 Chunk server N Bring computation directly to the data! Chunk servers also serve as compute servers Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 10 Task: Word Count Case 1: ■ File too large for memory, but all pairs fit in memory Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 11 Task: Word Count Case 1: ■ File too large for memory, but all pairs fit in memory Case 2: ■ Count occurrences of words: ■ words(doc.txt) | sort | uniq -c where words takes a file and outputs the words in it, one per a line Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 11 Task: Word Count Case 1: ■ File too large for memory, but all pairs fit in memory Case 2: ■ Count occurrences of words: ■ words(doc.txt) | sort | uniq -c ■ where words takes a file and outputs the words in it, one per a line ■ Case 2 captures the essence of MapReduce ■ Great thing is that it is naturally parallelizable Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 11 MapReduce: Overview ■ Sequentially read a lot of data ■ Map: ■ Extract something you care about ■ Group by key: Sort and Shuffle ■ Reduce: ■ Aggregate, summarize, filter or transform ■ Write the result Outline stays the same, Map and Reduce change to fit the problem Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 12 Map Reduce: The Map Step Input key-value pairs Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 13 Map Reduce: The Map Step Input key-value pairs Intermediate key-value pairs V map > v A V Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 13 Map Reduce: The Map Step Input Intermediate key-value pairs key-value pairs A Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 13 Map Reduce: The Map Step Input key-value pairs A I I maP A Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 13 Intermediate key-value pairs Map Reduce: The Map Step Input key-value pairs Intermediate key-value pairs V V V V map > L V A V Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 13 Map Reduce: The Map Step Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 13 MapReduce: The Reduce Step Intermediate key-value pairs Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 14 MapReduce: The Reduce Step Intermediate key-value pairs Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 14 MapReduce: The Reduce Step Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 14 MapReduce: The Reduce Step Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 14 MapReduce: The Reduce Step Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 14 MapReduce: The Reduce Step Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 14 MapReduce: The Reduce Step Intermediate key-value pairs Key-value groups V / V V / V V Output key-value pairs V V V V V Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 14 More Specifically ■ Input: a set of key-value pairs ■ Programmer specifies two methods: ■ Map(k, v) -> * ■ Takes a key-value pair and outputs a set of key-value pairs ■ E.g., key is the filename, value is a single line in the file ■ There is one Map call for every (k,v) pair ■ Reduce(k\ *) -> * ■ All values v' with same key W are reduced together and processed in i/'order ■ There is one Reduce function call per unique key W Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 15 MapReduce: Word Counting The crew of the space shuttle Endeavor recently returned to Earth as ambassadors, harbingers of a new era of space exploration. Scientists at NASA are saying that the recent assembly of the Dextre bot is the first step in a long-term space-based man/mache partnership. "The work we're doing now - the robotics we're doing - - is what we're going to need.......................... Big document Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 16 MapReduce: Word Counting The crew of the space shuttle Endeavor recently returned to Earth as ambassadors, harbingers of a new era of space exploration. Scientists at NASA are saying that the recent assembly of the Dextre bot is the first step in a long-term space-based man/mache partnership. "The work we're doing now - the robotics we're doing - - is what we're going to need.......................... Big document Provided by the programmer (key, value) Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 16 MapReduce: Word Counting Provided by the programmer The crew of the space shuttle Endeavor recently returned to Earth as ambassadors, harbingers of a new era of space exploration. Scientists at NASA are saying that the recent assembly of the Dextre bot is the first step in a long-term space-based man/mache partnership. "The work we're doing now - the robotics we're doing - - is what we're going to need.......................... Big document (key, value) (key, value) Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 16 MapReduce: Word Counting The crew of the space shuttle Endeavor recently returned to Earth as ambassadors, harbingers of a new era of space exploration. Scientists at NASA are saying that the recent assembly of the Dextre bot is the first step in a long-term space-based man/mache partnership. "The work we're doing now - the robotics we're doing - - is what we're going to need.......................... Big document Provided by the programmer (key, value) (key, value) Provided by the programmer (key, value) Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 16 MapReduce: Word Counting The crew of the space shuttle Endeavor recently returned to Earth as ambassadors, harbingers of a new era of space exploration. Scientists at NASA are saying that the recent assembly of the Dextre bot is the first step in a long-term space-based man/mache partnership. "The work we're doing now - the robotics we're doing - - is what we're going to need.......................... Provided by the programmer Provided by the programmer Big document (key, value) (key, value) (key, value) Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 16 MapReduce: Word Counting The crew of the space shuttle Endeavor recently returned Jo^^ajtt^js^ ambassadors, harbingers of a new era of space exploration. Scientists at nasa are saying mat tne recent assembly of the Dextre bot is the first step in man/mache partnership. "The work we're doing now - the robotics we're doing - - is what we're going to need.......................... Provided by the programmer Provided by the programmer Big document (key, value) (key, value) (key, value) Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 16 MapReduce: Word Counting The crew of the space shuttle Endeavor recently returned Jo^^ajtJ^js^ ambassadors, harbingers of a new era of space exploration. Scientists at nasa are saying mat tne recent assembly of the Dextre bot is the first step in man/mache partnership. "The work we're doing now - the robotics we're doing - - is what we're going to need.......................... Provided by the programmer (of, i) (the, i) (space, 1) (shuttle, 1) (Endeavor, 1) (recently, 1) Provided by the programmer Big document (key, value) (key, value) (key, value) Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 16 MapReduce: Word Counting The crew of the space shuttle Endeavor recently returned Jo^^ajtt^js^ ambassadors, harbingers of a new era of space exploration. Scientists at nasa are saying mat tne recent assembly of the Dextre bot is the first step in man/mache partnership. "The work we're doing now - the robotics we're doing - - is what we're going to need.......................... Provided by the programmer Big document (key, value) Provided by the programmer (key, value) (key, value) Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 16 MapReduce: Word Counting The crew of the space shuttle Endeavor recently returned Jo^^ajtJ^js^ ambassadors, harbingers of a new era of space exploration. Scientists at nasa are saying mat tne recent assembly of the Dextre bot is the first step in man/mache partnership. "The work we're doing now - the robotics we're doing - - is what we're going to need.......................... Provided by the programmer (crew, 1) (crew, 1) (space, 1) Big document (key, value) (shuttle, 1) (recently, 1) (key, value) Provided by the programmer (key, value) Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 16 Word Count Using MapReduce Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 17 Word Count Using MapReduce map(key, value): // key: document name; value: text of the document for each word w in value: emit(w, 1) Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 17 Word Count Using MapReduce map(key, value): // key: document name; value: text of the document for each word w in value: emit(w, 1) reduce(key, values): // key: a word; value: an iterator over counts result = 0 for each count v in values: result += v emit(key, result) Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 17 Map-Reduce: A diagram Input MAP: Read input an produces a set key-value pai Big document Intermediate Reduce: Collect all valu belonging totl key and outpi Group by Key 0 Grouped kl:v?v,v,v k2:v k3:v.,v k5\y © ©(■ v if \ b © © r v if Output III, if if if ] f f f T ^) i^m^) (^m) (^m) f }f v if kl :v kl :v k2:v kl:v k3:vk4:v k4:vk5:v k4:v kl:v k3:v Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 18 Map-Reduce: In Parallel Map Task 1 t r Map Task 2 | r Map Task 3 1 i i (*) ; ; ; ; kl ;v kl ;v k2;v kl:v i i k3:v k4:v k4:v k5;v i i k4;v kl;v k3;v Partitioning Function Partitionin \ Function Petitioning Function _ _1 [__ Sort and Group k2:v k4-:\ .v.v k5:v V Reduce Task 1 Sort and Group kl .vyyy k3:v.v Reduce Task 2 All phases are distributed with many tasks doing the work Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 19 Map-Reduce ■ Programmer specifies: ■ Map and Reduce and input files ■ Workflow: Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 20 Map-Reduce Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 20 Map-Reduce Programmer specifies: ■ Map and Reduce and input files Workflow: ■ Read inputs as a set of key-value-pairs ■ Map transforms input kv-pairs into a new set of kV-pairs Map o Map 2 Shuffle Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 20 Map-Reduce Programmer specifies: ■ Map and Reduce and input files Workflow: ■ Read inputs as a set of key-value-pairs ■ Map transforms input kv-pairs into a new set of kV-pairs ■ Sorts & Shuffles the kV-pairs to output nodes Map o 1— ■ Map 2 II Shuffle Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 20 Map-Reduce Programmer specifies: ■ Map and Reduce and input files Workflow: ■ Read inputs as a set of key-value-pairs ■ Map transforms input kv-pairs into a new set of kV-pairs ■ Sorts & Shuffles the kV-pairs to output nodes ■ All kV-pairs with a given k' are sent to the same reduce Map o Map 2 Shuffle Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 20 Map-Reduce Programmer specifies: ■ Map and Reduce and input files Workflow: ■ Read inputs as a set of key-value-pairs ■ Map transforms input kv-pairs into a new set of kV-pairs ■ Sorts & Shuffles the kV-pairs to output nodes ■ All kV-pairs with a given k' are sent to the same reduce ■ Reduce processes all kV-pairs grouped by key into new k'V'-pairs Map o Map 2 Shuffle Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 20 Map-Reduce Programmer specifies: Map and Reduce and input files Workflow: Read inputs as a set of key-value-pairs Map transforms input kv-pairs into a new set of kV-pairs Sorts & Shuffles the kV-pairs to output nodes All kV-pairs with a given k' are sent to the same reduce Reduce processes all kV-pairs grouped by key into new k'V'-pairs Write the resulting pairs to files Map o Map 2 Shuffle Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 20 Map-Reduce Programmer specifies: Map and Reduce and input files Workflow: Read inputs as a set of key-value-pairs Map transforms input kv-pairs into a new set of kV-pairs Sorts & Shuffles the kV-pairs to output nodes All kV-pairs with a given k' are sent to the same reduce Reduce processes all kV-pairs grouped by key into new k"v"-pairs Write the resulting pairs to files All phases are distributed with many tasks doing the work Map o Map 2 Shuffle Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 20 Data Flow ■ Input and final output are stored on a distributed file system (FS): ■ 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 Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 21 Coordination: Master ■ Master node takes care of coordination: ■ 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 Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 22 Dealing with 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 ■ Reduce task is restarted ■ Master failure ■ MapReduce task is aborted and client is notified Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 23 How many Map and Reduce jobs? ■ M map tasks, R reduce tasks ■ Rule of a thumb: ■ Make M much larger than the number of nodes in the cluster ■ One DFS chunk per map is common ■ Improves dynamic load balancing and speeds up recovery from worker failures ■ Usually R is smaller than M ■ Because output is spread across R files Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 24 Task Granularity & Pipelining ■ Fine granularity tasks: map tasks » machines ■ Minimizes time for fault recovery ■ Can do pipeline shuffling with map execution ■ Better dynamic load balancing Process Time----------------------> User Program Master MapReduceQ ... wait... Assign tasks to worker machines... Worker 1 Map 1 Map 3 Worker 2 Map 2 Worker 3 Read 1.1 Read 1.3 Read 1.2 1 Reduce 1 Worker 4 Read 2.1 Read 2.2 Read 2.3 Reduce 2 Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 25 Refinements: Backup Tasks ■ Problem ■ Slow workers significantly lengthen the job completion time: ■ Other jobs on the machine ■ Bad disks ■ Weird things ■ Solution ■ Near end of phase, spawn backup copies of tasks ■ Whichever one finishes first "wins" ■ Effect ■ Dramatically shortens job completion time Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 26 Refinement: Combiners Often a Map task will produce many pairs of the form (k,vj, (Km?), ... for the same key k ■ E.g., popular words in the word count example Can save network time by pre-aggregating values in the mapper: ■ combine(k, list(v-i)) v2 ■ Combiner is usually same as the reduce function Works only if reduce function is commutative and associative Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 27 Refinement: Combiners Back to our word counting example: ■ Combiner combines the values of all keys of a single mapper (single machine): — o (A3) (A.C) (A.D) u ) Combiner \ (B.2) (CI) (0.2) (E.1) (B.1) J ♦ Combiner (D.2) (A.2) (C.I) (B.1) Shuffle (A.2) (B.3) (C2) (0.4) (E.1) Much less data needs to be copied and shuffled! Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 28 Refinement: Partition Function ■ Want to control how keys get partitioned ■ Inputs to map tasks are created by contiguous splits of input file ■ Reduce needs to ensure that records with the same intermediate key end up at the same worker ■ System uses a default partition function: ■ hash(key) mod R ■ Sometimes useful to override the hash function: ■ E.g., hash(hostname(URL)) mod Rensures URLs from a host end up in the same output file Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 29 Problems Suited for Map-Reduce Example: Host size ■ Suppose we have a large web corpus ■ Look at the metadata file ■ Lines of the form: (URL, size, date,...) ■ For each host, find the total number of bytes ■ That is, the sum of the page sizes for all URLs from that particular host ■ Other examples: ■ Link analysis and graph processing ■ Machine Learning algorithms Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 31 Example: Language Model ■ Statistical machine translation: ■ Need to count number of times every 5-word sequence occurs in a large corpus of documents ■ Very easy with MapReduce: ■ Map: ■ Extract (5-word sequence, count) from document ■ Reduce: ■ Combine the counts Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 32 Example: Join By Map-Reduce ■ Compute the natural join R(A,B) M S(B,C) ■ R and S are each stored in files ■ Tuples are pairs (a,b) or (b,c) Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 33 Map-Reduce Join ■ Use a hash function h from B-values to l...k ■ A Map process turns: ■ Each input tuple R(a,b) into key-value pair (b,(a,R)) ■ Each input tuple S(b,c) into (b,(c,S)) ■ Map processes send each key-value pair with key b to Reduce process h(b) ■ Hadoop does this automatically; just tell it what k is. ■ Each Reduce process matches all the pairs (b,(a,R)) with all (b,(c,S)) and outputs (a,b,c). Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 34 Cost Measures for Algorithms ■ In Map Reduce we quantify the cost of an algorithm using 1. Communication cost = total I/O of all processes 2. Elapsed communication cost = max of I/O along any path 3. (Elapsed) computation cost analogous, but count only running time of processes Note that here the big-0 notation is not the most useful (adding more machines is always an option) Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 35 Example: Cost Measures ■ For a map-reduce algorithm: ■ Communication cost = input file size + 2 x (sum of the sizes of all files passed from Map processes to Reduce processes) + the sum of the output sizes of the Reduce processes. ■ Elapsed communication cost is the sum of the largest input + output for any map process, plus the same for any reduce process Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 36 What Cost Measures Mean ■ Either the I/O (communication) or processing (computation) cost dominates ■ Ignore one or the other ■ Total cost tells what you pay in rent from your friendly neighborhood cloud ■ Elapsed cost is wall-clock time using parallelism Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 37 Cost of Map-Reduce Join ■ Total communication cost = 0(|R| + |S| + |R N S|) ■ Elapsed communication cost = O(s) ■ We're going to pick k and the number of Map processes so that the I/O limit s is respected ■ We put a limit s on the amount of input or output that any one process can have, s could be: What fits in main memory ■ What fits on local disk ■ With proper indexes, computation cost is linear in the input + output size ■ So computation cost is like comm. cost Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 38 Pointers and Further Reading Implementations ■ Google ■ Not available outside Google ■ Hadoop ■ An open-source implementation in Java ■ Uses HDFS for stable storage ■ Download: http://lucene.apache.orq/hadoop/ ■ Aster Data ■ Cluster-optimized SQL Database that also implements MapReduce Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 40 Resources ■ Hadoop Wiki ■ Introduction ■ http://wiki.apache.org/lucene-hadoop/ ■ Getting Started ■ http://wiki.apache.org/lucene-hadoop/GettingStartedWithHadoop ■ Map/Reduce Overview ■ http://wiki.apache.org/lucene-hadoop/HadoopMapReduce ■ http://wiki.apache.org/lucene-hadoop/HadoopMapRedClasses ■ Eclipse Environment ■ http://wiki.apache.org/lucene-hadoop/EclipseEnvironment ■ Javadoc ■ http://lucene.apache.org/hadoop/docs/api/ Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 41 Resources ■ Releases from Apache download mirrors ■ http://www.apache.org/dvn/closer.cgi/lucene/had pop/ ■ Nightly builds of source http://people.apache.org/dist/lucene/hadoop/nig htlyi ■ Source code from subversion ■ http://lucene.apache.org/hadoop/version control .html Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 42