Distributed architectures for big data processing and analytics (2023/2024)
Courses

Distributed architectures for big data processing and analytics (2023/2024)

General Information

SSD: ING-INF/05

CFU: 8

Professor: Paolo Garza

Teaching Assistant: Simone Papicchio

Teaching Material

Introduction
  • Introduction to the course content and exam rules (pdf)
  • Introduction to Big Data (pdf)
  • Big Data Architectures (pdf)
Hadoop and MapReduce
  • Introduction to Apache Hadoop and the MapReduce programming paradigm (pdf)
    • Interaction with HDFS and Hadoop using the command line (pdf)
  • Hadoop implementation of MapReduce (pdf)
    • Source code of the Word Count Ecplise project (WordCount.zip) – Use the import maven project option to import it into Visual Studio Code
    • BigData@Polito environment + Jupyter – How to submit MapReduce jobs on BigData@Polito (pdf)
  • MapReduce – Design patterns – Part 1 (pdf)
  • MapReduce and Hadoop – Advanced Topics: Multiple inputs, Multiple outputs, Distributed cache (pdf)
  • MapReduce – Design patterns – Part 2 (pdf)
  • MapReduce – Relational Algebra/SQL operators (pdf)
Spark
  • Introduction to Apache Spark (pdf)
    • How to submit Spark applications (pdf)
    • How to use Jupyter Notebooks for your Spark applications (pdf)
    • You can install PySpark and JupyterLab using Conda/Miniconda/pip (instructions here)
  • RDD-based programs
    • RDDs: creation, basic transformations and actions (pdf)
    • Key-value RDDs: transformations and actions on key-value RDDs (pdf)
      • Inner join, left outer join, right outer join, full outer join, and “NOT IN” with PairRDDs: Examples – Notebook (JoinsRDD.zip) – Uploaded on April 21, 2024
    • DoubleRDDs (pdf)
    • Advanced Topics: Cache, accumulators, broadcast variables, custom partitioners, broadcast join (pdf)
  • Spark SQL and DataFrames
  • Data mining and Machine learning algorithms with Spark MLlib
  • GraphX/GraphFrames
    • Introduction to GraphX and GraphFrames (pdf)
    • Graph Algorithms with GraphFrames (pdf)
      • Simple example – Jupyter notebook (GraphFrameExamples.zip)
      • Select kernel GraphFrames (Yarn) to run it on jupyter.polito.it
      • Run “pyspark –packages graphframes:graphframes:0.8.1-spark3.0-s_2.12 –repositories https://repos.spark-packages.org” to run it locally on your PC – Use package graphframes:graphframes:0.8.0-spark2.4-s_2.11 if you locally installed Spark 2 instead of Spark 3
  • Streaming data analytics

Exercises

MapReduce
  • MapReduce Exercises (slides)
  • How to configure Visual Studio Code on your personal laptop: 📘guide.
    • Note that you must also install JDK 1.8 and select it for the imported project inside the IDE. If you have already installed the JDK environment but the version is greater than JDK 1.8, you must also install JDK 1.8.

    • Windows users only: You must configure the winutils (🗃️winutils.zip) and set up some environmental variables. Follow this 📘extra guide for the complete configuration.

    • There are multiple versions of the basic projects. The version with libraries is the only one you can use on the LABINF computers. Use it on your laptop if you are not interested in running the applications locally. All the other versions are Maven projects, so you can use them locally on your personal laptop to write the code and then run it locally inside Visual Studio Code or on the BigData@Polito cluster. The legend is as follows: 📚lib: Project/template with libraries, 🐧mavU: Maven project for Linux/MacOS, 🪟mavW: Maven project for Windows (Hadoop projects only).

    • Basic project for MapReduce applications (📚lib, 🐧mavU, 🪟mavW)
Spark

Laboratory Material

Team 1: Students from A to L – Tuesday from 11:30 to 13:00 (First lab activity – March 12, 2024) @ LABINF
Team 2: Students from M to Z – Friday from 11:30 to 13:00 (First lab activity – March 15, 2024) @ LABINF

  • How to configure Visual Studio Code on your personal laptop: 📘guide.
    • Note that you must also install JDK 1.8 and select it for the imported project inside the IDE. If you have already installed the JDK environment but the version is greater than JDK 1.8, you must also install JDK 1.8.
    • Windows users only: You must configure the winutils (🗃️winutils.zip) and set up some environmental variables. Follow this 📘extra guide for the complete configuration.
    • Laboratory materials are available in multiple versions. The version with libraries is the only one you can use on the LABINF computers. Use it on your laptop if you are not interested in running the applications locally. All the other versions are Maven projects, so you can use them locally on your personal laptop to write the code and then run it locally inside Visual Studio Code or on the BigData@Polito cluster. The legend is as follows: 📚lib: Project/template with libraries, 🐧mavU: Maven project for Linux/MacOS, 🪟mavW: Maven project for Windows (Hadoop projects only).
    • Basic project for MapReduce applications (📚lib, 🐧mavU, 🪟mavW)
  • How to configure JDK 1.8 on MAC in case of errors with standard procedure:
  • PySpark Installation Guide: How to run PySpark applications on your PC or Google Colab: You can install PySpark and JupyterLab using Conda/Miniconda/pip (instructions here)

Problem specifications/Lab solutions

Problem specification and input dataSolution (Maven-based)
Lab 1: Hadoop and MapReduce
Problem specification (pdf)
Basic project and small example dataset (Lab1_BigData_with_libraries_vscode.zip)
Basic project based on Maven – Use this version to run the MapReduce application locally on your own PC (DO NOT USE IT AT LABINF)
— Linux and macOS (Lab1.zip)
— Windows (Lab1_Windows.zip)
Bigger dataset: finefoods_text.txt (zip)		Solution: Bonus track Lab1_SolBonusMvn.zip
Lab 2: Filter with Hadoop MapReduce
Problem specification (pdf)
Skeleton project Hadoop — MapReduce (Lab2_Skeleton_with_libraries_vscode.zip)
Basic project based on Maven — Use this version of the project to run the MapReduce application locally on your own PC (DO NOT USE IT AT LABINF)
— Linux and macOS (Lab2_Skeleton.zip)
— Windows (Lab2Windows_Skeleton.zip)
Outputs of the first lab (OutputFolderLab1.zip) (OutputFolderLab1BonusTrack.zip). You can use them to test your application locally on your own PC if you are using Maven		Solution: Lab2_Sol.zip
Solution Bonus track: Lab2_SolBonus.zip
Lab 3: Frequently bought/reviewed together with Hadoop and MapReduce
Problem specification (pdf)
Skeleton project Hadoop — MapReduce (Lab3_Skeleton_with_libraries_vscode.zip)
Sample data (AmazonTransposedDataset_Sample.txt)
Basic project based on Maven — Use this version of the project to run the MapReduce application locally on your own PC (DO NOT USE IT AT LABINF)
— Linux and macOS (Lab3_Skeleton.zip)
— Windows (Lab3Windows_Skeleton.zip)		Solution: Lab3_DBD_Sol.zip – This project is based on mvn
— Comments on the three uploaded solutions (pdf)
— The second solution MUST NOT BE USED – It is highly inefficient
Lab 4: Normalized ratings for product recommendations with Hadoop MapReduce
Problem specification (pdf)
Skeleton Eclipse project Hadoop – MapReduce (Lab4_DBD_with_libraries.zip)
Basic project based on Maven – Use this version to run the MapReduce application locally on your own PC (DO NOT USE THIS ON LABINF PCs)
— Linux and macOS (Lab4_DBD_mvn.zip)
— Windows (Lab4_DBD_Windows_mvn.zip)
Sample file (ReviewsSample.csv)		Solution: Lab4_Sol.zip
Lab 5: Filter data and compute basic statistics with Apache Spark
Problem specification (pdf)
Sample file (SampleLocalFile.csv)		Solution: Lab5_DBD_Sol.zip
— Jupyter notebook (Lab5_Sol.ipynb)
— Python script (Lab5_Sol.py)
Lab 6: Frequently bought/reviewed together application with Apache Spark
Problem specification (pdf)
Sample dataset (ReviewsSample.csv)		Solution: Lab6_DBD_Sol.zip
— Jupyter notebook (Lab6_Sol.ipynb)
— Python script (Lab6_Sol.py)
Lab 7: Bike sharing data analysis
Problem specification (pdf)
Sample data (zip)
Example KML file (zip)
KML file containing the result of the analysis setting the threshold to 0.6 and running the program on the HDFS file (zip)		Solution: Lab7_DBD_Sol.zip
— Jupyter notebook (Lab7_Sol.ipynb)
— Python script (Lab7_Sol.py)
Lab 8: Bike sharing data analysis based on Spark SQL
Problem specification (pdf)
Sample data (zip)		Solution Lab8_DBD_Sol.zip
— Jupyter notebooks (Lab8_Sol.ipynb and Lab8_SolSQL.ipynb)
— Python scripts (Lab8_Sol.py and Lab8_SolSQL.py)
Lab 9: A classification pipeline with MLlib + SparkSQL
Problem specification (pdf)
Sample data (zip)		Solution Lab9_DBD_Sol.zip
— Jupyter notebooks
Lab10: GraphFrame
Problem specification (pdf)
Data (zip)		Solution Lab10_DBD_Sol.zip
— Jupyter notebooks – Updated on May 30, 2024. distinct() has been added in Task 2.
Lab11: Tweet analysis – Spark streaming
Problem specification (pdf)
Example files – tweets (zip)		Solution Lab11_DBD_Sol.zip
— Jupyter notebooks

Previous exam examples

ExamsSolutions
Exam September 6, 2024 (pdf)Question 1: (a) – The three codes are equivalent. They are based on commutative functions/methods.
Question 2: (a) – There are 3 distinct keys emitted by the map phase. Hence, the reduce method is invoked 3 times. It follows that the sum of the values of the three instances of numCitiesD is 3.
MapReduce and Spark (DBD_Exam20240906Sol.zip)
Exam July 19, 2024 (pdf)Question 1: (b) – 2 times – Three actions are based on the content of the input file, but highTempRDD is cached. Hence, the input file is read once to compute the value of the count action applied to tempRDD and then one more time to compute the content of highTempRDD, which is then used to calculate the results of the actions count and reduce applied to highTempRDD. Globally, due to the cache of highTempRDD, the input file is read twice.
Question 2: (d) – 6 – There are 6 input lines => the map method is invoked, overall, 6 times.
MapReduce and Spark (DBD_Exam20240719Sol.zip)
Exam July 5, 2024 (pdf)Question 1: (c) – Application B is not equivalent to A and C because .reduce(lambda v1,v2: min(v1, v2) ).filter(lambda value : value>5) is not equivalent to .filter(lambda value : value>5).reduce(lambda v1,v2: min(v1, v2) ). The two functions are not commutative.
Question 2: (a) – Considering all instances of the reducer class, the reduce method is invoked 3 times overall (2 + 1 + 0).
MapReduce and Spark (DBD_Exam20240705Sol.zip)
Sketch of a solution based on SQL (SQLBasedSolution.pdf)
Exam February 20, 2024 (pdf)Question 1: (a), Question 2: (b)
MapReduce and Spark (DBD_Exam20240220Sol.zip) – Uploaded on June 16, 2024
Exam September 18, 2023 (pdf)Question 1: (c), Question 2: (c)
MapReduce and Spark (Paper-based sketch of the solution – No code_ Exam20230918.pdf) – Uploaded on July 4, 2024
Exam July 19, 2023 (pdf)Question 1: (a), Question 2: (b)
MapReduce and Spark (DBD_Exam20230719Sol.zip) – Updated on June 9, 2024, with an SQL-based solution and some example data
Exam June 26, 2023 (pdf)Question 1: (b), Question 2: (c)
MapReduce and Spark (DBD_Exam20230626Sol.zip) – Updated on June 8, 2024, with an SQL-based solution and some example data
Exam September 1, 2022 (pdf)Question 1: (b), Question 2: (d)
MapReduce and Spark (DBD_Exam20220901Sol.zip)
Exam July 18, 2022 (pdf)Question 1: (b), Question 2: (b)
MapReduce and Spark (DBD_Exam20220718Sol.zip) – Updated on June 9, 2024, with an SQL-based solution – Example related to “static windows” and how to manage them either RDD or Spark SQL APIs
Exam June 27, 2022 (pdf)Question 1: (c), Question 2: (a)
MapReduce and Spark (DBD_Exam20220607Sol.zip) – Updated on June 8, 2024, with an SQL-based solution and some example data
Exam February 10, 2022 (pdf)Question 1: (a), Question 2: (b)
MapReduce and Spark (DBD_Exam20220210Sol.zip)
Exam September 17, 2021 (pdf)Question 1: (b), Question 2: (a)
MapReduce and Spark (DBD_Exam20210917.zip)
Exam July 5, 2021 (pdf)Question 1: (c), Question 2: (a)
MapReduce and Spark (DBD_Exam20210705Sol.zip) – Updated on May 7, 2024, with an SQL-based solution
Exam June 21, 2021 (pdf)Question 1: (b), Question 2: (a)
MapReduce and Spark (DBD_Exam20210621Sol.zip)
Exam July 20, 2020 (pdf)Question 1: (d), Question 2: (b)
Question 2 – Note that there are three actions. Hence, the input file is read three times.
MapReduce and Spark (DBD_Exam20200720Sol.zip)
Exam June 27, 2020 (pdf)Question 1: (b), Question 2: (a)
MapReduce and Spark (DBD_Exam20200627Sol.zip)
More examples of multiple choice questions (pdf)
		Question 1: (c)
Question 2: (d)
Question 3: (d)
Question 4: (d)
Question 5: (b)
Question 6: (d)
GraphFrame – Examples of multiple choice questions (pdf)Question 1: (d)
Question 2: (c)

Additional material

Slides and screencasts about Java (kindly provided by Prof. Torchiano) (link)
Focus on the following subset of slides/lectures (for students who have never used Java):
— OO Paradigm and UML (The UML part is not mandatory)
— The Java Environment
— Java Basic Features
— Java Inheritance