Download - Spark Driven Big Data Analytics

Spark Driven Big Data Analytics

Inosh GoonewardenaAssociate Technical Lead - WSO2, Inc.

Agenda● What is Big Data?

● Big Data Analytics

● Introduction to Apache Spark

● Apache Spark Components & Architecture

● Writing Spark Analytic Applications

What is Big Data?● Big data is a term for data sets that are so large and complex in nature

● Constitute structured, semi-structured and unstructured data

● Big Data cannot easily be managed by traditional RDBMS or statistics tools

Characteristics of Big Data - The 3Vs

http://itknowledgeexchange.techtarget.com/writing-for-business/files/2013/02/BigData.001.jpg

Sources of Big Data● Banking transactions

● Social Media Content

● Results of scientific experiments

● GPS trails

● Financial market data

● Mobile-phone call detail records

● Machine data captured by sensors connected to IoT devices

● ……..

Traditional Vs Big Data

Attribute Traditional Data Big Data

Volume Gigabytes to Terabytes Petabytes to Zettabytes

Organization Centralized Distributed

Structure Structured Structured, Semi-structured & Unstructured

Data Model Strict schema based Flat schema

Data Relationship Complex interrelationships Almost flat with few relationships

Big Data Analytics● Process of examining large data sets to uncover hidden patterns, unknown

correlations, market trends, customer preferences and other useful business information.

● Analytical findings can lead to better more effective marketing, new revenue opportunities, better customer service, improved operational efficiency, competitive advantages over rival organizations and other business benefits.

http://searchbusinessanalytics.techtarget.com/definition/big-data-analytics

4 types of Analytics● Batch

● Real-time

● Interactive

● Predictive

Challenges of Big Data Analytics● Traditional RDBMS fail to handle Big Data

● Big Data cannot fit in the memory of a single computer

● Processing of Big Data in a single computer will take a lot of time

● Scaling with traditional RDBMS is expensive

Traditional Large-Scale Computation● Traditionally, computation has been processor-bound

○ Relatively small amounts of data○ Significant amount of complex processing performed on that data

● For decades, the primary push was to increase the computing power of a single machine

○ Faster processor, more RAM

Hadoop● Hadoop is an open source, Java-based programming framework that

supports the processing and storage of extremely large data sets in a distributed computing environment

http://bigdatajury.com/wp-content/uploads/2014/03/030114_0817_HadoopCoreC120.png

The Hadoop Distributed File System - HDFS● Responsible for storing data on the cluster

● Data files are split into blocks and distributed across multiple nodes in the cluster

● Each block is replicated multiple times○ Default is to replicate each block three times○ Replicas are stored on different nodes○ This ensures both reliability and availability

MapReduce● MapReduce is the system used to process data in the Hadoop cluster

● A method for distributing a task across multiple nodes

● Each node processes data stored on that node - Where possible

● Consists of two phases:○ Map - process the input data and creates several small chunks of

data○ Reduce - process the data that comes from the mapper and

produces a new set of output

● Scalable, Flexible, Fault-tolerant & Cost effective

MapReduce - Example

http://www.cs.uml.edu/~jlu1/doc/source/report/img/MapReduceExample.png

Limitations of MapReduce● Slow due to replication, serialization, and disk IO

● Inefficient for:○ Iterative algorithms (Machine Learning, Graphs & Network

Analysis)○ Interactive Data Mining (R, Excel, Adhoc Reporting, Searching)

Apache Spark● Apache Spark is a cluster computing platform designed to be fast and

general-purpose

● Extends the Hadoop MapReduce model to efficiently support more types of computations, including interactive queries and stream processing

● Provides in-memory cluster computing that increases the processing speed of an application

● Designed to cover a wide range of workloads that previously required separate distributed systems, including batch applications, iterative algorithms, interactive queries and streaming

Features of Spark● Speed − Spark helps to run an application in Hadoop cluster, up to 100

times faster in memory, and 10 times faster when running on disk.

● Supports multiple languages − Spark provides built-in APIs in Java, Scala, or Python.

● Advanced Analytics − Spark not only supports ‘Map’ and ‘reduce’. It also supports SQL queries, Streaming data, Machine learning (ML), and Graph algorithms.

Spark Stack

https://media.licdn.com/mpr/mpr/shrinknp_800_800/AAEAAQAAAAAAAAb5AAAAJDUwZGZmZWEwLWNhZGEtNDc4NC1hOTkyLTVjMTNiYmUzNjVmNw.png

Components of Spark● Apache Spark Core − Underlying general execution engine for spark

platform that all other functionality is built upon. Provides In-Memory computing and referencing datasets in external storage systems.

● Spark SQL − Component on top of Spark Core that introduces a new data abstraction called SchemaRDD, which provides support for structured and semi-structured data.

● Spark Streaming − Leverages Spark Core's fast scheduling capability to perform streaming analytics. Ingests data in mini-batches and performs RDD transformations on those mini-batches of data.

Components of Spark● MLlib − Distributed machine learning framework above Spark. Provides

multiple types of machine learning algorithms, including binary classification, regression, clustering and collaborative filtering, as well as supporting functionality such as model evaluation and data import.

● GraphX − Distributed graph-processing framework on top of Spark. Provides an API for expressing graph computation that can model the user-defined graphs by using Pregel abstraction API.

Why a New Programming Model?● MapReduce simplified big data analysis.

● But users quickly wanted more:○ More complex, multi-pass analytics (e.g. ML, graph)○ More interactive ad-hoc queries○ More real-time stream processing

● All 3 need faster data sharing in parallel apps

Data Sharing in MapReduce● Iterative Operations on MapReduce

● Interactive Operations on MapReduce

https://www.tutorialspoint.com/apache_spark/images/iterative_operations_on_mapreduce.jpg

https://www.tutorialspoint.com/apache_spark/images/interactive_operations_on_mapreduce.jpg

Data Sharing using Spark RDD● Iterative Operations on Spark RDD

● Interactive Operations on Spark RDD

https://www.tutorialspoint.com/apache_spark/images/iterative_operations_on_spark_rdd.jpg

https://www.tutorialspoint.com/apache_spark/images/interactive_operations_on_spark_rdd.jpg

Execution Flow

http://cdn2.hubspot.net/hubfs/323094/Petr/cluster-overview.png?t=1478804099651

Execution Flow (contd.)1. A standalone application starts and instantiates a SparkContext instance.

Once the SparkContext is initiated the application is called the driver.2. The driver program ask for resources to launch executors from the cluster

manager.3. The cluster manager launches executors.4. The driver process runs through the user application. Depending on the

actions and transformations over RDDs task are sent to executors.5. Executors run the tasks and save the results.6. If any worker crashes, its tasks will be sent to different executors to be

processed again.

Terminology● Application

○ User program built on Spark. Consists of a driver program and executors on the cluster.

● Application Jar○ A jar containing the user's Spark application and its dependencies

except Hadoop & Spark Jars

● Driver Program○ The process where the main method of the program runs○ Runs the user user code that creates a SparkContext, creates

RDDs, and performs actions and transformation

Terminology (contd.)● SparkContext

○ Represents the connection to a Spark cluster○ Driver programs access Spark through a SparkContext object ○ Can be used to create RDDs, accumulators and broadcast

variables on that cluster

● Cluster Manager○ An external service to manage resources on the cluster

(standalone manager, YARN, Apache Mesos)

Terminology (contd.)● Deploy Mode

○ cluster - driver inside the cluster○ client - driver outside the cluster

● Worker node○ Any node that can run application code in the cluster

● Executor○ A process launched for an application on a worker node, that runs

tasks and keeps data in memory or disk storage across them. Each application has its own executors.

Terminology (contd.)● Task

○ A unit of work that will be sent to one executor

● Job○ A parallel computation consisting of multiple tasks that gets

spawned in response to a Spark action (e.g. save, collect).

● Stage○ Smaller set of tasks that each job is divided.○ Sequential and depend on each other

Spark PillarsTwo main abstractions of Spark.

● RDD - Resilient Distributed Dataset

● DAG - Direct Acyclic Graph

RDD (Resilient Distributed Dataset)● Fundamental data structure of Spark

● Immutable distributed collection of objects

● The data is partitioned across machines in the cluster that can be operated in parallel

● Fault-tolerant

● Support two types of operations○ Transformation○ Action

RDD - Transformations● Returns a pointer to new RDD

● lazily evaluated

● Step in a program telling Spark how to get data and what to do with it.

● Some of the most popular Spark transformations:○ map○ filter○ flatMap○ groupByKey○ reduceByKey

RDD - Actions● Return a value to the driver program after running a computation on the

dataset

● Some of the most popular Spark actions:○ reduce○ collect○ count○ take○ saveAsTextFile

DAG (Direct Acyclic Graph)A Graph that doesn’t link backwards. Defines sequence of computations performs on data.

DAG (Direct Acyclic Graph)


A failure happens at one of the operation


Replay and reconstruct the RDD

Writing Analytics Applications1. SparkConf conf = new SparkConf().setMaster("local").setAppName("Word Count App");2. JavaSparkContext sc = new JavaSparkContext(conf);3. 4. JavaRDD<String> lines = spark.read().textFile(args[0]).javaRDD();5.6. JavaRDD<String> words = lines.flatMap(new FlatMapFunction<String, String>() {7. @Override8. public Iterator<String> call(String s) {9. return Arrays.asList(s.split(" ")).iterator();10. }11. });12.13. JavaPairRDD<String, Integer> ones = words.mapToPair(new PairFunction<String, String, Integer>() {14. @Override15. public Tuple2<String, Integer> call(String s) {16. return new Tuple2<>(s, 1);17. }18. });19.20. JavaPairRDD<String, Integer> counts = ones.reduceByKey(new Function2<Integer, Integer, Integer>() {21. @Override22. public Integer call(Integer i1, Integer i2) {23. return i1 + i2;24. }25. });26.27. List<Tuple2<String, Integer>> output = counts.collect();28 for (Tuple2<?,?> tuple : output) {29 System.out.println(tuple._1() + ": " + tuple._2());30. }

Writing Analytics Applications (contd.)SparkConf conf = new SparkConf().setMaster("local").setAppName("Word Count App");JavaSparkContext sc = new JavaSparkContext(conf);

JavaRDD<String> lines = spark.read().textFile(args[0]).javaRDD();

JavaRDD<String> words = lines.flatMap(new FlatMapFunction<String, String>() { @Override public Iterator<String> call(String s) { return Arrays.asList(s.split(" ")).iterator(); }});

JavaPairRDD<String, Integer> ones = words.mapToPair(new PairFunction<String, String, Integer>() { @Override public Tuple2<String, Integer> call(String s) { return new Tuple2<>(s, 1); }

});

JavaPairRDD<String, Integer> counts = ones.reduceByKey(new Function2<Integer, Integer, Integer>() { @Override public Integer call(Integer i1, Integer i2) { return i1 + i2; }

});

List<Tuple2<String, Integer>> output = counts.collect();for (Tuple2<?,?> tuple : output) {

System.out.println(tuple._1() + ": " + tuple._2());}

Green colored codes are not executing on the driver. Those are executing on another node.

Writing Analytics Applications (contd.)SparkConf conf = new SparkConf().setMaster("local").setAppName("Word Count App");JavaSparkContext sc = new JavaSparkContext(conf);

JavaRDD<String> lines = spark.read().textFile(args[0]).javaRDD();

JavaRDD<String> words = lines.flatMap(new FlatMapFunction<String, String>() { @Override public Iterator<String> call(String s) { return Arrays.asList(s.split(" ")).iterator(); }});

JavaPairRDD<String, Integer> ones = words.mapToPair(new PairFunction<String, String, Integer>() { @Override public Tuple2<String, Integer> call(String s) { return new Tuple2<>(s, 1); }

});

JavaPairRDD<String, Integer> counts = ones.reduceByKey(new Function2<Integer, Integer, Integer>() { @Override public Integer call(Integer i1, Integer i2) { return i1 + i2; }

});

List<Tuple2<String, Integer>> output = counts.collect();for (Tuple2<?,?> tuple : output) {

System.out.println(tuple._1() + ": " + tuple._2());}

Blue colored codes are transformations. Red colored code is an action. Green colored code is stuffs that happens locally (not Spark related things)

Writing Analytics Applications (Java 8)

SparkConf conf = new SparkConf().setMaster("local").setAppName("Word Count App");JavaSparkContext sc = new JavaSparkContext(conf);

// Load the input data, which is a text file read from the command lineJavaRDD<String> lines = sc.textFile(filename);

JavaRDD<String> words = lines.flatMap(line -> Arrays.asList(line.split(" ")).iterator());

JavaPairRDD<String, Integer> ones = words.mapToPair(w -> new Tuple2<>(w, 1));

JavaPairRDD<String, Integer> counts = ones.reduceByKey((i1, i2) -> i1 + i2);

List<Tuple2<String, Integer>> output = counts.collect();for (Tuple2<?,?> tuple : output) { System.out.println(tuple._1() + ": " + tuple._2());}

Questions?