Machine Learning - Spring 2022 - Class 20

  • hadoop video: https://www.youtube.com/watch?v=irK7xHUmkUA&list=PLAwxTw4SYaPkXJ6LAV96gH8yxIfGaN3H-&index=20
       - videos 20 to 35

  • quick review of hadoop basics
       - consists of two key components
          - HDFS: Hadoop Distributed File System
             - allows us to store large amounts of data
             - distributes the files across multiple computers
             - stores redundant copies of the data for reliability
          - MapReduce framework
             - computational framework to allow programmers to analyze very large data sets
             - splits up the computation across multiple computers
             - tries to take advantage of data locality based on the HDFS
             - two key phases: map phase and reduce phase
             
       - these two components are what make up a hadoop cluster

       - from a programmer's perspective, there are a number of other tools
          - there is an extensive set of libraries that dictate how you interact with the system
             - we'll see some of these as time goes
          - there are a number of programs that allow us to interact with the cluster, e.g.
             - interact with the HDFS
             - submit jobs to the system
             - check the status of the system
             - ...

  • Virtual machines
       - A virtual machine is a simulation of a computer (and all its parts) on another machine
       - Very common-place now
       - We're going to use a VM to run our hadoop cluster
          - This simplifies things so everyone has their own instance of hadoop running
          - Make installation *much* easier
          - Only downside is that you won't really see major speedups

  • HDFS basics
       - it is only a file system
          - in particular there is no operating system, no shell, etc.
       - it is stored within the local file system
       - it is a distributed file system so it is distributed across the local file systems of all of the computers in the cluster
          - this is what allows it to store very large data sets
       - to interact with it you use a call to hadoop, specifically:

          > hdfs dfs <some_command>

       - because there is no shell, there is no current working directory, therefore
          - no cd
          - all commands either
             - require the full path, that is starting with /
             - or are assumed to start in the users home directory

  • HDFS file structure
       - like the normal linux file structure, everything starts at /
       - instead of having "home" directories, user files go in a directory called "/user"
       - the default location for interactions with the HDFS is:
          /user/<username>

          for our VM, the username is "training", so all the files will go in:

          /user/training/

          - you can see all the user directories by typing

          > hdfs dfs -ls /user/
          Found 3 items
          drwxr-xr-x - hue supergroup 0 2013-09-05 20:08 /user/hive
          drwxr-xr-x - hue hue 0 2013-09-10 10:37 /user/hue
          drwxr-xr-x - training supergroup


  • HDFS commands
       - HDFS supports many of the basic file management commands
       - You can see a list of the dfs commands online at: https://images.linoxide.com/hadoop-hdfs-commands-cheatsheet.pdf or simply by typing:
          > hdfs dfs


  • HDFS basic commands
       - -ls: lists the files in the directory
          - by default, it just gives an ls of your user directory:   

          > hdfs dfs -ls

          - alternatively, you can specify a path and it will give an ls of that directory

          > hdfs dfs -ls /user
          
       - -ls -R: lists the file in the directory AND recurses and lists all subdirectories
          - the default input to many hadoop program is a directory containing a collection of files, so this can be useful
          - given that you can do cd and some of the other associated tricks with a traditional terminal, this helps

          > hdfs dfs -ls -R /user/
          drwxr-xr-x - hue supergroup 0 2013-09-05 20:08 /user/hive
          drwxrwxrwx - hue supergroup 0 2013-09-05 20:08 /user/hive/warehouse
          drwxr-xr-x - hue hue 0 2013-09-10 10:37 /user/hue
          drwxrwxrwt - hue hue 0 2013-09-10 10:39 /user/hue/jobsub
          drwx--x--x - training hue 0 2013-09-10 10:37 /user/hue/jobsub/_training_-design-1
          -rw-r--r-- 1 training hue 677 2013-09-10 10:37 /user/hue/jobsub/_training_-design-1/workflow.xml
          drwx--x--x - training hue 0 2013-09-10 10:39 /user/hue/jobsub/_training_-design-2
          -rw-r--r-- 1 training hue 1054 2013-09-10 10:39 /user/hue/jobsub/_training_-design-2/workflow.xml
          drwxr-xr-x - training supergroup 0 2016-11-03 20:48 /user/training


       - -mv: moves files WITHIN the HDFS

       - -cp: similar to mv, copies file WITHIN the HDFS

       - -rm: removes a particular file

       - -mkdir: makes a directory WITHIN the HDFS
          > hdfs -mkdir temp
          > hdfs dfs -ls
          Found 1 items
          drwxr-xr-x - training supergroup 0 2016-11-03 20:50 temp

       - -rm -R
          - similar to rm -r on normal file system
          - deletes the directory, the files in that directory AND recurses
              - be careful!

       - -chmod and -chown: same commands as on the normal file system for handing permissions and ownership



  • Putting and getting data to the HDFS
       - so far, we've only talked about how to move files around, copy them, etc.
       - because it's a separate file system, there are also special commands for moving files from the current computer's file system TO the HDFS and vice versa
       
       - putting files onto the HDFS
          - to add files (e.g. files that you want your programs to process) you need to put them onto the HDFS
          - -put
             > hdfs dfs -put file1.txt temp
             (copies file1.txt from the local file system TO temp directory, actually /user/training/temp/ directory, on the HDFS)

             - notice that the first argument is a file/directory on the local file system and the second argument is a location on the HDFS
             
             > hdfs dfs -ls temp   
             Found 1 items
             -rw-r--r-- 1 training supergroup 57 2016-11-03 20:53 temp/file1.txt

       - viewing file on the HDFS
          - if you just want to peek at the contents of a file on the HDFS there are couple of commands you can use
          - -cat: display the contents of a file (same as cat on normal fs)
             > hdfs dfs -cat temp/file1.txt
             this is the first file
             it has lots of good stuff in it

          - -tail: display the last 1K of the file
             - if the file is very large, you may not want to see the entire thing
             - you can use tail to peak just at the last bit of the file

       - getting files from the HDFS
          - eventually, you may want to actually get files from the HDFS
          - -get: get the specified file OR directory to the local file system
             > hdfs dfs -get temp/file1.txt .
             (copies file1.txt into my current directory)

             > hdfs dfs -get temp .
             (copies the directory and all contents to the current directory)

             - notice that the first argument is a file/directory on the HDFS and the second argument is the location on the local file system

          - -getmerge
             - a common output of a mapreduce program is a directory filled with different files, each representing a portion of the final solution
             - the getmerge function allows us to grab the files in a directory and merge them into a single file on the local file system
             > hdfs dfs -getmerge temp temp.txt
             (copies all of the files in temp into a single file, temp.txt)

       
  • Interacting with the HDFS programmatically
       - You can also interact with the HDFS programmatically
          - we won't cover much of this in this class, however, there are many examples online

  • general mapreduce program pipeline
       1. the input is provided to the mapreduce program
          - think of the input as a giant list of elements
          - elements are ALWAYS some key/value pairs
          - however, the default key/value pair is:
             - value = line in a file
             - key = byte offset into the file
       2. each key/value pair is passed to the mapping function
          - the mapping function takes a key/value pair as input
          - does some processing
          - and outputs one or more key/value pairs as output (not necessarily the same types as input)
       3. all of the output pairs are grouped by the key
          - this results in: key -> value1, value2, value3, ... for all the values associated with that specific key
          - this is still a key value pair
             - the key = key
             - the value = and iterator over values
       4. these new key/value pairs are then passed to the reducer function
          - input is key -> value iterator
          - does some processing (often some sort of aggregation)
          - outputs the final key/value pairs, which should the the answer (or at least answer to the subproblem)

  • writing a mapreduce program
       - three components:
          1. map function
          2. reduce function
          3. driver

  • word count
       - let's write a program that takes a file as input and produces a list of all the words in the file and the number of times that word occurs in the file

       - general overview: first, let's look at how we can break this down into a map step and a reduce step
          - map step for word count?
             - input is a line
             - two output options
                - option 1: word -> # of occurrence in this line
                - option 2: word -> 1 for each word in the line
             - either of the options is fine, however, most often will choose option 2
                - simpler to implement
                - you want the map function to be as fast and as simple as possible
                - you want to avoid having to declare/create new objects since this takes time
                   - remember that this map function is going to be called for *every* line in the data
                - you want the processing for each call to map to be as consistent as possible
          - reduce step
             - the reduce step gets as input the output from the map step with the values aggregated into an iterator per key
                - in our case: word -> 1, 1, 1, 1, 1, 1, 1, 1 (an iterator with a bunch of 1s)
             - all we need to do is some these up and output a pair of word -> sum

          - driver
             - the driver just sets a bunch of configuration parameters required to get the job going