SparkPaper

Spark: Unified Framework to manage Big Data
Abstract
Apache Spark is an open source big data framework to build around speed, ease of use and analytics. It
was originally developed in 2009 in UC Berkeley’s AMPLab, and open sourced in 2010 as an Apache
project. Spark is complete framework for big data. Spark comes packaged with higher-level libraries,
including support for SQL queries, streaming data, machine learning and graph processing. These
standard libraries increase developer productivity and can be seamlessly combined to create complex
workflows. Spark enables applications in Hadoop clusters to run up to 100 times faster in memory and 10
times faster even when running on disk.

Introduction
Over the past decade the era of information world has evolving towards the data science, the way of
collecting data has been tremendously changes over the period of time. Historically data was being
generated an accumulated by workers, company hire employee to enter the user data into the computer
and then internet comes along than user can generate their own data which we may called the year of
social network, now a days machines are accumulating data. So while collecting those huge amounts of
data a single computer didn’t handle those data, infinite number of server and parallel processing need to
manage those data so the distributed computer system comes along.
Within the past 11 year in big data Hadoop is the only choice and it proven to be the solution to store and
manage those huge amount of data. It is an open source project manages by Apache Software Foundation
for distributed storage and processing of large data set. It uses the Hadoop distributed file system known
as HDFS which is able to store large files across the multiple servers.
Hadoop doesn’t meet the requirement of today’s need it doesn’t maintain the speed in calculating the
large amount of data set also it will take more time to run queries.
Spark was introduced by Apache Software foundation which is a fast, in-memory engine of data
processing. It is a 10x faster on disk and 100x faster in memory than Hadoop. Although it use the Hadoop
file system to store the data but it is not the modified version of Hadoop. Spark can run standalone along
with Hadoop. It has its own clustering system it only use Hadoop as storage.

Apache Spark is a powerful open source project written in Scala Programming Language for processing
large data set with speed, ease of use, and sophisticated analytics. It provides in-memory analytics which
is faster than Hadoop/Hive. It designed for running iterative algorithms and interactive analytics. Spark
use Resilient Distributed Datasets (RDD) for its data structure which is an immutable object. While
computing in different nodes each dataset in RDD is divided into logical partitions. The RDD is a fault-
tolerant collection of elements that can operate on in parallel because of it RDD know how to recreate and
recomputed the datasets. RDD supports in-memory processing computation which store the state of
memory as an object across the jobs and the object is sharable between those jobs.
Figure 1: Spark Components
There are additional libraries which is part of the spark other than the spark core that gives more
capabilities in machine learning and big data.
Spark Streaming: It is used to process the data in real time typically based on micro batch style of
computing and processing. Apache Spark has provided an unified engine that natively supports both batch

and streaming workloads. This is different from other systems that either have a processing engine
designed only for streaming, or have similar batch and streaming APIs but compile internally to different
engines. Spark’s single execution engine and unified programming model for batch and streaming lead to
some unique benefits over other traditional streaming systems
Spark SQL: It provides the JDBC API and allows running SQL like queries on Spark. Spark SQL is a
component on top of Spark Core that introduces a new data abstraction called SchemaRDD, which
provides support for structured and semi-structured data.
Spark MLlib: It consists the algorithms and utilities including regression, clustering, classification and
other machine learning. MLlib is a distributed machine learning framework above Spark because of the
distributed memory-based Spark architecture. It is, according to benchmarks, done by the MLlib
developers against the Alternating Least Squares (ALS) implementations.
Spark GraphX: It includes a collection of graph algorithm and build graph analytics tasks. GraphX is a
distributed graph-processing framework on top of Spark. It provides an API for expressing graph
computation that can model the user-defined graphs by using Pregel abstraction API.

Spark Program Flow
In cluster spark runs independent processes by the SparkContext object in main program which is called
driver program. The applications that are in different types of cluster will connect to SparkContext. It
connects the application that occupied the resource of clusters. To run the computation and store data of
the application it will executes on nodes in the clusters and then it sends application code to the executors.
At the end SparkContext sends tasks to the executors to run.
Figure 2 Spark Component

Hadoop MapReduce Vs. Spark
Hadoop and Spark are both different and independent big-data frameworks. Hadoop is a distributed data
infrastructure whereas Spark is a data-processing tool that operates those data, It doesn’t do distributed
storage.
Hadoop comes with storage known as Hadoop Distributed File System and processing module called
MapReduce so we can say that it is a complete package, in other hand spark doesn’t come with the
storage it needs to be connected with the file system, it can use HDFS or other file system. Initially Spark
was designed to enhance the Hadoop so that many recommend using HDFS.
Spark is lot faster than MapReduce because it processes the data in memory instead of disk that
MapReduce use. MapReduce read data from cluster, perform the task, write result in cluster, read from
cluster and again doing the same thing but Spark completes its full task in-memory and real-time. Read
data from cluster, perform the task in memory, write back to cluster So that it is 100 times faster than
Hadoop.
Spark provides multi language api such as Java, Scala, Python so that developers can quickly develop
application and build parallel apps. Along with Spark supports SQL queries, streaming data, and complex
analytics along with map and reduce.

Spark can runs on Hadoop, Mesos, standalone or in the cloud. It supports multiple data sources such as
HDFS, Cassandra, HBase, S3.
The following is the major cases Spark over Hadoop
 Iterative Algorithms in Machine Learning
 Interactive Data Mining and Data Processing
 Spark is a fully Apache Hive-compatible data warehousing system that can run 100x faster than
Hive.
 Stream processing: Log processing and Fraud detection in live streams for alerts, aggregates and
analysis
 Sensor data processing: Where data is fetched and joined from multiple sources, in-memory
dataset really helpful as they are easy and fast to process.
Upon first glance, it seems that using Spark would be the default choice for any big data application.
However, that’s not the case. MapReduce has made inroads into the big data market for businesses that
need huge datasets brought under control by commodity systems. Spark’s speed, agility, and relative ease
of use are perfect complements to MapReduce’s low cost of operation.
The truth is that Spark and MapReduce have a symbiotic relationship with each other. Hadoop provides
features that Spark does not possess, such as a distributed file system and Spark provides real-time, in-
memory processing for those data sets that require it. The perfect big data scenario is exactly as the
designers intended—for Hadoop and Spark to work together on the same team.

Spark Hadoop
Apache Spark processes data in-memory Hadoop MapReduce persists back to the disk after a
map or reduces action.
Spark can do great in real-time processing as
well as batch processing.
Hadoop Map Reduce is great for batch processing. If
you want a real-time option you’ll need to use another
platform like Storm or Impala, and for graph
processing you can use Graph.
Spark is generally a lot faster than MapReduce,
10 times faster than Map Reduce for batch
processing and up to 100 times faster for in-
memory analytics.
Map Reduce is slow
Spark performs better when all the data fits in
the memory.
Hadoop MapReduce is designed for data that doesn’t
fit in the memory
The process will start from beginning if process
crashes in the middle of execution.
The process will continue where it is left off if the
process crashes in the middle of execution.
Batch Processing Real-time processing
Stores Data on Disk Stores Data in Memory
Written in Java Written in Scala
Less secured because it is still in early stage of
development.
More secured , has security features like kerborose,
sentry etc.
How to install Spark
Step 1: Verify Java Installation
First of all you need to verify Java installation which is mandatory for installing Spark. You can verify it
by using command
Java –version
In case you have not install Java then do it.
Step 2: verifying Scala installation
You can verify Scala installation by using this command
Scala –version
In case you don’t have Scala installed then proceed to next step.
Step 3: Downloading Apache Spark
Download the latest version of Spark https://spark.apache.org/downloads.html and copy the Spark and
the same Scala folder and extract it.

Step 4: Installation Spark
Setting up the environment for Spark
Setx JAVA_HOME
Setx c:Program FilesJavajdk1.8.0_71
Step 5: Verifying the Spark Installation
You can use command: spark –shell
If spark is installed successfully then you get out like this:
Spark assembly has been built with Hive, including Datanucleus jars on classpath
Using Spark's default log4j profile: org/apache/spark/log4j-defaults.properties
15/06/04 15:25:22 INFO SecurityManager: Changing view acls to: hadoop
15/06/04 15:25:22 INFO SecurityManager: Changing modify acls to: hadoop
15/06/04 15:25:22 INFO SecurityManager: SecurityManager: authentication disabled;
ui acls disabled; users with view permissions: Set(hadoop); users with modify permissions: Set(hadoop)
15/06/04 15:25:22 INFO HttpServer: Starting HTTP Server
15/06/04 15:25:23 INFO Utils: Successfully started service 'HTTP class server' on port 43292.
Welcome to
____ __
/ __/__ ___ _____/ /__
_ / _ / _ `/ __/ '_/
/___/ .__/_,_/_/ /_/_ version 1.4.0
/_/
Using Scala version 2.10.4 (Java HotSpot(TM) 64-Bit Server VM, Java 1.7.0_71)
Type in expressions to have them evaluated.

Spark context available as sc
scala>
Word Count Examples
Step 1 : Load text file
scala> val input = sc.textFile("Z:BDAhomework5obamaspeech1.txt")
Step 2: Use lambda expression to count the grouping text.
scala> val wordCounts = input.flatMap(line=>line.split("
")).map(word=>(word,1)).reduceByKey((a,b)=>a+b)
Note: With single line of lambda code it can group the text and count the frequency.
Step 3: Collect and Count
scala> wordCounts.collect()
res9: Array[(String, Int)] = Array((segregation,,1), (ills,1), (country,1), (God,3), (greater,4), (patrol,1), (responsibility,1), (carried,3), (filled,1),
(Now,,1), (prosperous,,1), (play,,1), (been,7), (wage,,1), (met.,1), (citizenship.,1), (challenges,2), (over,3), (packed,1), (Guided,1), (make,2),
(conflict,1), (tribe,1), (prefer,1), (long,,1), (threat,,1), (month,1), (real.,1), (are,22), (economy,3), (yet,,1), (indifference,1), (scarcely,1), (raise,1),
(safely,1), (sacrifices,1), (aims,1), (our,57), (plans.,1), (storms,1), (use;,1), (dust,1), (values,1), (alliances,1), (presidential,1), (come.,1),
(Afghanistan.,1), (ancestors.,1), (pursue,1), (Yet,,1), (them,4), (coldest,1), (chosen,1), (sapping,1), (village,1), (pleasures,1), (firm,1),
(alongside,1), (ourselves,,1), (pass;,1), (cha.
Spark with R Studio
Sys.setenv(SPARK_HOME="C:/Users/suraj/Downloads/spark")
.libPaths(c(file.path(Sys.getenv("SPARK_HOME"),"R","lib"),.libPaths()))
library(SparkR)
sc <-sparkR.init(master="local")
sqlContext <-sparkRSQL.init(sc)
DF <-createDataFrame(sqlContext,faithful)
head(DF)
> head(DF)
eruptions waiting
1 3.600 79
2 1.800 54
3 3.333 74
4 2.283 62
5 4.533 85
6 2.883 55
localDF <-data.frame(name=c("John","Smith","Sarah"),ages=c(19,23,18))
df<-createDataFrame(sqlContext,localDF)
printSchema(df)
path <- file.path(Sys.getenv("SPARK_HOME"), "examples/src/main/resources/people.json")
peopleDF <- jsonFile(sqlContext, path)
printSchema(peopleDF)

registerTempTable(peopleDF, "people")
# SQL statements can be run by using the sql methods provided by sqlContext
teenagers <- sql(sqlContext, "SELECT name FROM people WHERE age >= 13 AND age <= 19")
teenagersLocalDF <- collect(teenagers)
print(teenagersLocalDF)
name
1 Justin
sparkR.stop()
Conclusion
Apache Spark is the sparkly new evolving tool in Big Data. Spark has excellent performance and is highly
cost-effective. It’s compatible with all Hadoop’s data sources and files formats and has friendly APIs that
are available in different languages. It also includes graph processing and machine learning capabilities.
It does have a collection of all required tools to work with Big Data and also support third party tool for
storage and other processes.

References
1. What is Apache Spark
https://databricks.com/spark/about
2. Is Apache Spark going to replace Hadoop
http://aptuz.com/blog/is-apache-spark-going-to-replace-hadoop/
3. Big Data Processing with Apache Spark – Part 1: Introduction
http://www.infoq.com/articles/apache-spark-introduction
4. Apache Spark RDD
http://www.tutorialspoint.com/apache_spark/apache_spark_rdd.htm
5. Spark SQL & DataFrames
http://spark.apache.org/sql/
6. An Overview of Apache Spark
https://www.youtube.com/watch?v=mL5dQ_1gkiA

SparkPaper

More Related Content

What's hot (20)

Viewers also liked (11)

Similar to SparkPaper (20)

SparkPaper