Getting started in Apache Spark and Flink (with Scala) - Part II

11.07.2016 | Spark tutorial | A. Panchenko, G. Hintz, S. Remus
Getting started in Apache Spark
and Flink (with Scala)
Alexander Panchenko, Gerold Hintz, Steffen Remus

11.07.2016 | Spark tutorial | A. Panchenko, G. Hintz, S. Remus | 2
Outline
 Scala
 basics of Scala programming language
 Spark
 motivation / what do you get on top of MapReduce
 basics of Spark: RDDs, transformations, actions, shuffling
 “tricks” useful in Spark context
 Spark Hands-on session
 run Spark notebook and solve easy tasks
 setup Spark project & submit job to cluster
 Flink
 theory
 difference from Spark

Three main benefits to use Spark
1. Spark is easy to use—you can develop applications on your laptop,
using a high-level API
2. Spark is fast, enabling interactive use and complex algorithms
3. Spark is a general engine, letting you combine multiple types of
computations (e.g., SQL queries, text processing, and machine
learning) that might previously have required different engines.
This tutorial is based on the book by creators of Spark:
Karau H., Konwinski A., Windell P., Zaharia M. “Learning
Spark. Lighting-fast Data Analysis.” O’Really. 2015

Data Science Tasks
Experimentation: development of the model
 Python, MATLAB, R
 iPython notebooks
 Interactive computing
 Easy-to-use
Production: using the model
 Java, Scala, C++/C
 Unit tests
 Fault tolerance
 No interactive computing
 Scalability
Scala + Spark can be used for both!

A Brief History of Spark
 Spark is an open source project
 Spark started in 2009 as a research project in the UC Berkeley RAD Lab
 Research papers were published about Spark at academic conferences
and soon after its creation in 2009
 In 2011, the AMPLab started to develop higher-level components on
Spark, such as Shark (Hive on Spark) and Spark Streaming
 Currently one of the most active project in Scala language:

What Is Apache Spark?
 Spark Core: resilient distributed dataset (RDD)
 Spark SQL: Hive tables, Parquet, JSON, Datasets

What Is Apache Spark?
 Components for distributed execution in Spark

Spark Runtime Architecture
The components of a distributed Spark application

Spark Runtime Architecture
 The master/slave architecture with one central coordinator and many
distributed workers
 The central coordinator is called the driver
 The driver communicates with distributed workers called executors
 The driver is the process where the main() method of your program runs
 The driver:
 Converting a user program into tasks
 Scheduling tasks on executors

Downloading Spark and Getting Started
 Download a version “Pre-built for Hadoop 2.X and later”:
http://spark.apache.org/downloads.html
 Directories you see here that come with Spark:
 README.md
 Contains short instructions for getting started with Spark.
 bin
 Contains executable files that can be used to interact with Spark in various
ways (e.g., the Spark shell, which we will cover later in this chapter).
 core, streaming, python, …
 Contains the source code of major components of the Spark project.
 examples
 Contains some helpful Spark standalone jobs that you can look at and run to
learn about the Spark API.

Introduction to Spark’s Scala Shell
 Run: bin/spark-shell
 Type in the shell the Scala line count:
 We can run parallel operations on the RDD, such as counting the lines of
text in the file or printing the first one

Filtering: lambda functions
 Filtering example (Scala):
 Filtering example (Java 7):
 Filtering example (Java 8):

Standalone Spark Applications
 Link to Spark (Maven or SBT), e.g.:
 Write a sample class, e.g. word count:

Standalone Spark Applications
 SBT build file
 Build JAR and run it:

Programming with RDDs
RDD -- Resilient Distributed Dataset
 Immutable distributed collection of objects
 Each RDD is split into multiple partitions
 Partitions may be computed on different nodes
Creating an RDD
 Loading an external dataset
 Distributing a collection of objects

Programming with RDDs
 Once created, RDDs offer two types of operations:
 Transformations
 Actions
 Transformations construct a new RDD from a previous one
 Actions, compute a result based on an RDD
 either return it to the driver program
 or save it to an external storage system, e.g. HDFS
 RDDs are recomputed each time you run an action
 To reuse an RDD you need to persist it in memory:

Spark Execution Steps (Shell & Standalone)
1. Create some input RDDs from external data.
2. Transform them to define new RDDs using transformations like filter().
3. Persist any intermediate RDDs that will need to be reused.
4. Launch actions such as count() and first() to kick off a parallel
computation, which is then optimized and executed by Spark.

RDD Operations: Transformations
 filter() operation does not mutate the existing inputRDD
 It returns a pointer to an entirely new RDD
 inputRDD can still be reused later in the program, e.g.:

RDD Operations: Actions
Return some result and launch actual computation:
 take() to retrieve a small number of elements

Common Transformations and Actions
Element-wise transformations
 Mapped and filtered RDD from an input RDD:
 Squaring the values in an RDD:

Element-wise transformations
 Splitting lines into multiple words:
 Difference between flatMap() and map() on an RDD:

Some simple set operations:

Basic RDD transformations on an RDD containing {1, 2, 3, 3}:

Two-RDD transformations on RDDs containing {1, 2, 3} and {3, 4, 5}:

Basic actions on an RDD containing {1, 2, 3, 3}:

Persistence (Caching)
Double execution: Reusing result:
 Persistence levels:

Working with Key/Value Pairs
 Pair RDDs are a useful building block in many programs
 Allow you to act on each key in parallel or regroup data
 For instance:
 reduceByKey() method that can aggregate data for each key
 join() method that can merge two RDDs by grouping elements with the same
key
 Creating Pair RDDs = creating Scala tuples:
 Creating a pair RDD using the first word as the key

Transformations on Pair RDDs
 Transformations on one pair RDD (example: {(1, 2), (3, 4), (3, 6)})

 Transformations on one pair RDD (example: {(1, 2), (3, 4), (3, 6)})

 Transformations on two pair RDDs (rdd = {(1, 2), (3, 4), (3, 6)}
other = {(3, 9)})

 Using partial functions syntax for Pair RDDs in Scala
 Simple filter on second element:

Word and document counts:
 Per-key average with reduceByKey() and mapValues():

Word count example revisited:

Example of a join (inner join is the default):

Actions Available on Pair RDDs
Actions on pair RDDs (example ({(1, 2), (3, 4), (3, 6)}))

Example: PageRank
 links – (pageID, link List) – a list of neighbors of each page
 ranks – (pageID,rank) – current rank for each page

Important topics not covered in this intro
MLlib
 Machine Learning in the distributed way
 Basic Linear Algebra in the distributed way: sparse and dense vectors
and matrices
Partitioning
 No free lunch, neither automagic scaling of any algorithm
 Making efficient algorithm = trying to minimize shuffling of the data
Spark SQL, Spark 2.0, Datasets, DataFrames
 Something like Python’s pandas or R’s DataFrame
 Great for interactive data mining and for working with CSV files

Getting started in Apache Spark and Flink (with Scala) - Part II

More Related Content

What's hot

Viewers also liked

Similar to Getting started in Apache Spark and Flink (with Scala) - Part II

More from Alexander Panchenko

Recently uploaded

Getting started in Apache Spark and Flink (with Scala) - Part II