Read Carbondata Table from Apache Hive

Apache Carbondata an indexed columnar data store heavily depends on Apache Spark but also supports other Big Data frameworks like Apache Hive and Presto. This article explains how to read a Carbondata table created in Apache Spark from Apache Hive in two sections: 1. How to create a table in HDFS using Apache Spark, 2. How to read the Carbondata table from Apache Hive.

Requirements:

• Oracle JDK 1.8
• Apache Spark
• Apache Hadoop (Carbondata officially support Hive 2.x. In this article, Apache Hadoop 2.7.7 is used)
• Apache Hive (Carbondata officially support Hive 2.x. So better to stick to 2.x version. In this article, Apache Hive 2.3.6 is used to demonstrate the integration)
• Carbondata libraries

Please follow the Integrate Carbondata with Apache Spark Shell article to compile Carbondata from source and integrate it with Apache Spark. This article is written based on the assumption that you have already followed all the steps from the above-mentioned article.

Read More

Integrate Carbondata with Apache Spark Shell

Apache Carbondata an indexed columnar data store solution for fast analytics on big data platform, e.g.Apache Hadoop, Apache Spark, etc. This article is written to provide a quick start guide on how to integrate Carbondata with Apache Spark Shell. Why another article while there is a quick start guide on the official website? Things are not always as smooth as expected. In my experience, integrating Carbondata with Apache Spark using pre-built binaries didn't work as expected. So here is the quick start tutorial.

Requirements:
Carbondata requires Java 1.7 or 1.8 to run and Apache Maven to build from source. Please make sure that you have Oracle JDK 1.8, supporting Apache Maven and Git to setup Carbondata. If you don't have Oracle JDK or Apache Maven installed in your system, please follow the given links below to install them first.

• Install Oracle JDK 8 on Linux
• Install Apache Maven on Linux

Read More

Spark 06: Broadcast Variables

If you read the Spark 04: Key-Value RDD and Average Movie Ratings article, you might wonder what to do with popular movie IDs printed at the end. A data analyst cannot ask his/her users to manually check those IDs in a CSV file to find the movie name. In this article, you will learn how to map those movie IDs to movie names using Apache Spark's variable broadcasting.

Suppose you want to share a read-only data that can fit into memory with every worker in your Spark cluster, broadcast that data. The broadcasted variable will be distributed only once and cached in every worker node so that it can be reused any number of times. More about broadcasting will be covered later in this article after the code example.

Read More

Spark 05: List Action Movies with Spark flatMap

Welcome to the fifth article in the series of Apache Spark tutorials. In this article, you will learn the application of flatMap transform operation. After the introduction to flatMap operation, a sample Spark application is developed to list all action movies from the MovieLens dataset.

In the previous articles, we have used the map transform operation which transforms an entity into another entity where the transformation is one-to-one. For example, suppose you have a String RDD named lines, applying lines.map(x => x.toUpperCase) operation creates a new String RDD with the same number of records but with uppercase string literals as shown below:

Read More

Spark 04: Key-Value RDD and Average Movie Ratings

In the first article of this series: Spark 01: Movie Rating Counter, we created three RDDs (data, filteredData and ratingData) each contains a singular datatype. For example, data and filteredData were String RDDs and the ratingRDD was a Float RDD. However, it is common to use an RDD which can store complex datatypes especially Key-Value pairs depending on the requirement. In this article, we will use a Key-Value RDD to calculate the average rating of each movie in our MoviLens dataset.  Those who don't have the MovieLens dataset, please visit the Spark 01: Movie Rating Counter article to setup your environment.

As you already know, the ratings.csv file has the fields movieId and rating. A given movie may get different ratings from different users. To get the average ratings of each movie, we need to add all ratings of each movie individually and divide the sum by the number of ratings.

Read More

Spark 03: Understanding Resilient Distributed Dataset

You are not qualified as an Apache Spark developer until you know what is a Resilient Distributed Dataset (RDD). It is the fundamental technique to represent data in the Spark memory. There are advanced data representation techniques like DataFrame built on top of RDD. However, it is always better to start with the most basic dataset: RDD. RDD is nothing other than a data structure with some special properties or features.

We all know that Apache Spark is a distributed general-purpose cluster-computing framework. There are some common problems faced in a distributed environment including but not limited to:

1. Remote access of data is expensive
2. High chance of failure
3. Runtime errors are expensive and hard to track
4. Wasting computing power is way too expensive

RDD is designed to address the abovementioned problems. In the following section, you will see the properties of RDD and how it solves these problems.

Read More

Spark 02: Scala Cheat Sheet for Java Developers

This article introduces Scala to those Java developers who don't know Scala. I assume here that you already know Java (preferably Java 8) so that you can compare the features of Scala with Java. Please be informed that this article is not an end to end Scala tutorial. I am covering only the fundamentals of Scala which are used in my Apache Spark tutorials.

First of all, remember that Scala is a JVM based language which is running on top of your regular Java Virtual Machine. The whole purpose of Scala is providing a convenient functional programming language (at that time Java 8 wasn't there). However since Scala is built on top of Java, you can access Java libraries and API from your Scala code.

To play with Scala, please setup Scala on IntelliJ IDEA or install a command line Scala version in your system.

Read More

Spark 01: Movie Rating Counter

Apache Spark is a must to know framework for big data developers. This is an attempt to write a series of articles on Apache Spark to train you from zero to hero. In this series of articles, I will use the latest Apache Spark release which is 2.4.0 as of 2019 January. In the first few articles, we will code and test Apache Spark on IntelliJ Idea. As you may already know, Apache Spark is developed using Scala and of course, there are APIs available for other languages including Java and Python. Still, Scala has preferred over other languages for its performance and compact code. Therefore, you need to prepare the environment first.

My articles will be based on Frank Kane's course on Udemy: Apache Spark 2 with Scala - Hands On with Big Data! I highly recommend his tutorial if you prefer for a video tutorial.

Setup the Environment

Apache Spark 2.4.0 depends on Scala 2.12 which in turn depends on Java 1.8. Scala is unlike Java not very version compatible language. Therefore, please take special care on choosing versions.

Step 1:

Install Oracle Java Development Kit 1.8 on your system. Linux users can follow this article: Install Oracle JDK 8 on Linux

Step 2:

Install the latest IntelliJ IDEA. Again, Linux users can follow my article: Install IntelliJ IDEA on Linux

Step 3:
Install Scala plugin in IntelliJ IDEA. Regardless of your operating system, you can follow the article Setup Scala on IntelliJ IDEA.

Read More