Book Image

Hadoop Blueprints

By : Anurag Shrivastava, Tanmay Deshpande
Book Image

Hadoop Blueprints

By: Anurag Shrivastava, Tanmay Deshpande

Overview of this book

If you have a basic understanding of Hadoop and want to put your knowledge to use to build fantastic Big Data solutions for business, then this book is for you. Build six real-life, end-to-end solutions using the tools in the Hadoop ecosystem, and take your knowledge of Hadoop to the next level. Start off by understanding various business problems which can be solved using Hadoop. You will also get acquainted with the common architectural patterns which are used to build Hadoop-based solutions. Build a 360-degree view of the customer by working with different types of data, and build an efficient fraud detection system for a financial institution. You will also develop a system in Hadoop to improve the effectiveness of marketing campaigns. Build a churn detection system for a telecom company, develop an Internet of Things (IoT) system to monitor the environment in a factory, and build a data lake – all making use of the concepts and techniques mentioned in this book. The book covers other technologies and frameworks like Apache Spark, Hive, Sqoop, and more, and how they can be used in conjunction with Hadoop. You will be able to try out the solutions explained in the book and use the knowledge gained to extend them further in your own problem space.
Table of Contents (14 chapters)
Hadoop Blueprints
About the Authors
About the Reviewers

Hadoop platform tools

We have successfully run our first MapReduce program on Hadoop written in Java. The Hadoop ecosystem offers a rich set of tools to perform various activities on the Hadoop cluster. The rich tool set is a mix of open source tools and commercial tools available from vendors. The Hadoop ecosystem of tools improves continuously through the very active open source community and commercial vendors who actively contribute to improvements.

In this section, we cover some of the popular tools that we will use to build solutions from Chapter 2, A 360-Degree View of Customer, onwards in this book. A very comprehensive list of tools is available on a website known as Hadoop Illuminated at this URL:

Figure 22 Hadoop: The ecosystem of tools

In the following sections, we will cover a brief overview of the tool after having learned what HDFS, MapReduce v2 and YARN are capable of. In the upcoming chapters, we will cover the installation and usage of these tools as we start using them.

Data ingestion tools

Data ingestion is the process of reading data from a source outside the HDFS and loading them onto the HDFS for storage and further processing. The data can come from flat files, relational database management systems, and through live data feeds. Three common tools to ingest incoming data in Hadoop are as follows:

  • Sqoop: Hadoop usually coexists with other databases in the enterprise. Apache Sqoop is used to transfer the data between Hadoop and relational database systems or mainframe computers that are ubiquitous in enterprises of all sizes. Typically, you can use Sqoop to import the data from Oracle or SQL Server, transform it using MapReduce, or other tools such as Hive and Pig, and then export the data back to other systems. Sqoop internally uses MapReduce to import data into the Hadoop cluster. Relational database management systems are used to store operational data such as point of sale transactions, orders and customer master data. Apache Sqoop offers fast data loading in HDFS by running data loading tasks in parallel.

  • Flume: Often we have to load data coming from streaming data sources in HDFS. Streaming data sources delivers data in the form of events, which arrive at random or fixed time intervals.

    Flume is used to ingest large volumes of data coming from sources such as web logs, click stream logs, twitter feeds and livestock feeds in the form of events. In Flume, such data is processed as one event at a time. Flume allows you to build multi-hop flows where an event can travel through before it is stored in the HDFS. Flume offers reliable event delivery. Events are removed from a channel only if they are stored in the next channel, or persisted in a permanent data store such as the HDFS. The Flume channel manages the recoverability of the events in case of failure by using a durable channel that makes use of the local filesystem.

  • Kafka: Kafka is a messaging system based on a publish subscribe (pubsub) model of messaging commonly used by JMS and ActiveMQ. However, the design of Kafka is akin to a distributed, partitioned and replicated commit log service. Kafka maintains message feeds, which are classified using categories known as topics. Producers are the processes that publish the messages to the topics so that Consumers can consume them. In the pubsub model, a published message gets broadcast to all the consumers. Kafka also supports a queuing model in which a published message gets picked by exactly one consumer who is part of a consumer group.

    Between producers and consumers, Kafka acts as a broker running on a cluster of servers. Typical use cases for Kafka are log aggregation, website activity monitoring and stream processing. Despite its similarity to Flume, Kafka is a general purpose commit-log service suitable for Hadoop, but also for many other systems, while Flume is a solution designed for Hadoop alone. It is not uncommon to see Kafka and Flume working together to build a data ingestion flow for the HDFS.

Data access tools

With the help of data access tools, you can analyze the data stored on Hadoop clusters to gain new insights from the data. Data access tools help us with data transformation, interactive querying and advanced analytics. In this section, we will cover commonly used open source data access tools available from the Apache Software Foundation:


In addition to open source tools, several commercial tools are available from vendors such as Datameer, IBM, and Cloudera

  • Hive: SQL is a widely understood query language. Hive provides an SQL-like query language known as Hive Query Language(HQL). Though HQL is limited in features compared to SQL language, it is still very useful to developers who are familiar with running SQL queries on relational database management systems. A bigger group of database programmers can be engaged in the Hadoop ecosystem, using Hive. The Hive service breaks down HQL statements into MapReduce jobs, which execute on the Hadoop cluster like any other MapReduce job. From the Hive 1.2 release, we can take advantage of Apache Tez. Apache Tez is a new application framework built on top of YARN. Using Apache Tez instead of MapReduce solves some of the inefficiencies associated with the planning and execution of queries. This makes Hive perform faster.

  • Pig: Apache Pig is another tool for analyzing large data sets that consists of a high-level language for expressing data analysis programs, coupled with an infrastructure for evaluating these programs. Pig uses a special language called Pig Latin. This programming language uses high-level constructs to shield the programmers from the complexity of coding MapReduce programs in Java. Pig comes with its own command line shell. You can load HDFS files into Pig and perform various transformations, and then store the results on the HDFS. Pig translates transformation tasks into MapReduce jobs. Once Pig scripts are ready they can be run with the help of a scheduler without any manual intervention to perform routine transformations.

  • Hbase: Apache Hbase is a NoSQL database that works on top of the HDFS. Hbase is a distributed and non-relational column-oriented data store modeled after Google's Bigtable. It gives you random, real-time, read and write access to data, which is not possible in HDFS. It is designed to store billions of rows and millions of columns. Hbase uses HDFS to store data but it is not limited to HDFS alone. Hbase delivers low access latency for a small amount data from a very large data set.

  • Storm: Apache Storm is a distributed near real-time computation system. Similarly to how Hadoop provides a set of general primitives for doing batch processing, Storm provides a set of general primitives for performing real-time computation. It defines its workflows in Directed Acyclic Graphs (DAG's) called topologies. These topologies run until they are shut down by the user or until they encounter an unrecoverable failure. Storm can read and write data from the HDFS. In its original form, Storm did not natively run on Hadoop clusters but it used Apache Zookeeper and its own master/ minion worker processes to coordinate topologies with a master and worker state. Now Storm is also available on YARN, bringing real-time streamed data-processing capabilities in Hadoop.

  • Spark: Apache Spark is the newest kid on the block. It performs the memory processing of data stored in the HDFS. Hadoop is inherently a batch-oriented data processing system. In order to build a data pipeline, we have to read and write data several times on the HDFS. Spark addresses this issue by storing the data in memory, which makes low latency data processing possible after the initial load of data from the HDFS into the RAM. Spark provides an excellent model for performing iterative machine learning and interactive analytics. However, Spark also excels in some areas similar to Storm's capabilities, such as near real-time analytics and ingestion. Apache Spark does not require Hadoop to operate. However, its data parallel paradigm requires a shared file system for the optimal use of stable data. The stable source can range from Amazon S3, NFS, MongoDB or, more typically, HDFS. (Ballou, 2014). Spark supports a number of programming languages such as Java, Python, R, and Scala. Spark supports relational queries using a Spark SQL, machine learning with Mlib, graph processing with GraphX, and streaming with Spark Streaming.

Monitoring tools

Hadoop monitoring tools monitor its infrastructure and resources. This information is useful to maintain the quality of service as per the service level agreements, and monitor them over and under the utilization of deployed resources for financial charging and capacity management. Ambari is the most popular open source tool for Hadoop monitoring. Hadoop vendor Cloudera has a proprietary tool, called Cloudera Manager, which is also used to monitor Cloudera Hadoop installations


Cloudera Manager has been positioned as the Enterprise Hadoop Admin System by Cloudera. It helps in the deployment and management of Hadoop clusters. Monitoring is just a subset of many other features offered by it.

In this section, we will cover Ambari.

Apache Ambari is a monitoring tool for Hadoop. It also covers the provisioning and management of Hadoop clusters. Ambari has a REST interface that enables the easy integration of Hadoop provisioning, management, and monitoring capabilities to other applications in the enterprise. You can get an instant insight into the health of a running Hadoop cluster by looking at the web-based dashboard of Ambari (Zacharias, 2014). Ambari can collect the operational metrics of a Hadoop cluster for analysis later. Using the alert framework of Ambari, you can define the rules to generate alerts that you can act upon.

Figure 23 Ambari dashboard

Data governance tools

Data governance (DG) refers to the overall management of the availability, usability, integrity, and security of the data employed in an enterprise. The data governance architecture of Hadoop is evolving, and is still very underdeveloped. Presently, there is no comprehensive data governance within the Hadoop stack, and integration with the external governance frameworks is lacking.

Industry is responding to the need to build a comprehensive data-governance architecture, because Hadoop is becoming the central building block of big data processing systems in enterprises.

Apache Atlas: The aim of the Apache Atlas initiative is to deliver a comprehensive data governance framework. The goals of this initiative are:

  • Support for data classification

  • Support for centralized auditing

  • Proving a search and lineage function

  • Building a security and policy engine