Broadcast variables allow the programmer to stay a read-only variable cached on each machine rather than shipping a copy of it with tasks. They can be used, for example, to give every node, a copy of a large input dataset, in an efficient manner. Spark also attempts to distribute broadcast variables using efficient broadcast algorithms to reduce communication cost.

Spark actions are executed through a group of stages, separated by distributed “shuffle” operations. Spark automatically broadcasts the common knowledge needed by tasks within each stage.

The knowledge broadcasted this manner is cached in serialized form and is deserialized before running each task. This means that explicitly creating broadcast variables, is only useful when tasks across multiple stages need the equivalent knowledge or when caching the knowledge in deserialized form is important.

Broadcast variables are created from a variable v by calling SparkContext.broadcast(v). The broadcast variable is a wrapper around v, and its value can be accessed by calling the value method. The code given below shows this −

scala> val broadcastVar = sc.broadcast(Array(1, 2, 3))

Output −

broadcastVar: org.apache.spark.broadcast.Broadcast[Array[Int]] = Broadcast(0)

After the broadcast variable is created, it should be used instead of the value v in any functions run on the cluster, so that v is not shipped to the nodes more than once. In addition, the object v should not be modified after its broadcast, in order to ensure that all nodes get the equivalent value of the broadcast variable.

Accumulators

Accumulators are variables that are only “added” to through an associative operation and can therefore, be efficiently supported in parallel. They can be used to implement counters (as in MapReduce) or sums. Spark natively supports accumulators of numeric types, and programmers can add support for new types. If accumulators are created with a name, they will be displayed in Spark’s UI. This can be useful for understanding the progress of running stages (NOTE − this is not yet supported in Python).

An accumulator is created from an initial value v by calling SparkContext.accumulator(v). Tasks running on the cluster can then add to it using the add method or the += operator (in Scala and Python). However, they cannot read its value. Only the driver program can read the accumulator’s value, using its value method.

The code given below shows an accumulator being used to add up the elements of an array −

scala> val accum = sc.accumulator(0)




scala> sc.parallelize(Array(1, 2, 3, 4)).foreach(x => accum += x)

If you want to see the output of above code then use the subsequent command −

scala> accum.value

Output

res2: Int = 10

Numeric RDD Operations

Spark allows you to do different operations on numeric knowledge, using one of the predefined API methods. Spark’s numeric operations are implemented with a streaming algorithm that allows building the model, one element at a time.

These operations are computed and returned as a StatusCounter object by calling status() method.

The subsequent is a list of numeric methods available in StatusCounter.

S.No	Methods & Meaning
1	count() Number of elements in the RDD.
2	Mean() Average of the elements in the RDD.
3	Sum() Total value of the elements in the RDD.
4	Max() Maximum value among all elements in the RDD.
5	Min() Minimum value among all elements in the RDD.
6	Variance() Variance of the elements.
7	Stdev() Standard deviation.

If you would like to use only one of these methods, you’ll call the corresponding method directly on RDD.

So, this brings us to the end of blog. This Tecklearn ‘Deep dive into Advanced programming in Spark’ helps you with commonly asked questions if you are looking out for a job in Apache Spark and Scala and Big Data Developer. If you wish to learn Apache Spark and Scala and build a career in Big Data Hadoop domain, then check out our interactive, Apache Spark and Scala Training, that comes with 24*7 support to guide you throughout your learning period. Please find the link for course details:

https://www.tecklearn.com/course/apache-spark-and-scala-certification/

Apache Spark and Scala Training

About the Course

Tecklearn Spark training lets you master real-time data processing using Spark streaming, Spark SQL, Spark RDD and Spark Machine Learning libraries (Spark MLlib). This Spark certification training helps you master the essential skills of the Apache Spark open-source framework and Scala programming language, including Spark Streaming, Spark SQL, machine learning programming, GraphX programming, and Shell Scripting Spark. You will also understand the role of Spark in overcoming the limitations of MapReduce. Upon completion of this online training, you will hold a solid understanding and hands-on experience with Apache Spark.

Why Should you take Apache Spark and Scala Training?

The average salary for Apache Spark developer ranges from approximately $93,486 per year for Developer to $128,313 per year for Data Engineer. – Indeed.com
Wells Fargo, Microsoft, Capital One, Apple, JPMorgan Chase & many other MNC’s worldwide use Apache Spark across industries.
Global Spark market revenue will grow to $4.2 billion by 2022 with a CAGR of 67% Marketanalysis.com

What you will Learn in this Course?

Introduction to Scala for Apache Spark

What is Scala
Why Scala for Spark
Scala in other Frameworks
Scala REPL
Basic Scala Operations
Variable Types in Scala
Control Structures in Scala
Loop, Functions and Procedures
Collections in Scala
Array Buffer, Map, Tuples, Lists

Functional Programming and OOPs Concepts in Scala

Functional Programming
Higher Order Functions
Anonymous Functions
Class in Scala
Getters and Setters
Custom Getters and Setters
Constructors in Scala
Singletons
Extending a Class using Method Overriding

Introduction to Spark

Introduction to Spark
How Spark overcomes the drawbacks of MapReduce
Concept of In Memory MapReduce
Interactive operations on MapReduce
Understanding Spark Stack
HDFS Revision and Spark Hadoop YARN
Overview of Spark and Why it is better than Hadoop
Deployment of Spark without Hadoop
Cloudera distribution and Spark history server

Basics of Spark

Spark Installation guide
Spark configuration and memory management
Driver Memory Versus Executor Memory
Working with Spark Shell
Resilient distributed datasets (RDD)
Functional programming in Spark and Understanding Architecture of Spark

Playing with Spark RDDs

Challenges in Existing Computing Methods
Probable Solution and How RDD Solves the Problem
What is RDD, It’s Operations, Transformations & Actions Data Loading and Saving Through RDDs
Key-Value Pair RDDs
Other Pair RDDs and Two Pair RDDs
RDD Lineage
RDD Persistence
Using RDD Concepts Write a Wordcount Program
Concept of RDD Partitioning and How It Helps Achieve Parallelization
Passing Functions to Spark

Writing and Deploying Spark Applications

Creating a Spark application using Scala or Java
Deploying a Spark application
Scala built application
Creating application using SBT
Deploying application using Maven
Web user interface of Spark application
A real-world example of Spark and configuring of Spark

Parallel Processing

Concept of Spark parallel processing
Overview of Spark partitions
File Based partitioning of RDDs
Concept of HDFS and data locality
Technique of parallel operations
Comparing coalesce and Repartition and RDD actions

Machine Learning using Spark MLlib

Why Machine Learning
What is Machine Learning
Applications of Machine Learning
Face Detection: USE CASE
Machine Learning Techniques
Introduction to MLlib
Features of MLlib and MLlib Tools
Various ML algorithms supported by MLlib

Integrating Apache Flume and Apache Kafka

Why Kafka, what is Kafka and Kafka architecture
Kafka workflow and Configuring Kafka cluster
Basic operations and Kafka monitoring tools
Integrating Apache Flume and Apache Kafka

Apache Spark Streaming

Why Streaming is Necessary
What is Spark Streaming
Spark Streaming Features
Spark Streaming Workflow
Streaming Context and DStreams
Transformations on DStreams
Describe Windowed Operators and Why it is Useful
Important Windowed Operators
Slice, Window and ReduceByWindow Operators
Stateful Operators

Improving Spark Performance

Learning about accumulators
The common performance issues and troubleshooting the performance problems

DataFrames and Spark SQL

Need for Spark SQL
What is Spark SQL
Spark SQL Architecture
SQL Context in Spark SQL
User Defined Functions
Data Frames and Datasets
Interoperating with RDDs
JSON and Parquet File Formats
Loading Data through Different Sources

Scheduling and Partitioning in Apache Spark

Concept of Scheduling and Partitioning in Spark
Hash partition and range partition
Scheduling applications
Static partitioning and dynamic sharing
Concept of Fair scheduling
Map partition with index and Zip
High Availability
Single-node Recovery with Local File System and High Order Functions

Got a question for us? Please mention it in the comments section and we will get back to you.

579