HADOOP IN PRACTICE, 2nd Edition – An updated guide to handling some of the ‘trickier and dirtier aspects of Hadoop’ – #programming #bookreview


Hadoop in Practice, Second Edition

Alex Holmes

(Manning – paperback )


The Hadoop world has undergone some big changes lately, and this hefty, updated edition offers excellent coverage of a lot of what’s new. If you currently work with Hadoop and MapReduce or are planning to take them up soon, give serious consideration to adding this well-written book to your technical library. A key feature of the book is its “104 techniques.” These show how to accomplish practical and important tasks when working with Hadoop, MapReduce and their growing array of software “friends.”

The author, Alex Holmes, has been working with Hadoop for more than six years and is a software engineer, author, speaker, and blogger specializing in large-scale Hadoop projects.

His new second edition, he writes, “covers Hadoop 2, which at the time of writing is the current production-ready version of Hadoop. The first edition of the book covered Hadoop 0.22 (Hadoop 1 wasn’t yet out), and Hadoop 2 has turned the world upside-down and opened up the Hadoop platform to processing paradigms beyond MapReduce. YARN, the new scheduler and application manager in Hadoop 2, is complex and new to the community, which prompted me to dedicate a new chapter 2 to covering YARN basics and to discussing how MapReduce now functions as a YARN application.”

In the book, Holmes notes that “Parquet has also recently emerged as a new way to store data in HDFS—its columnar format can yield both space and time efficiencies in your data pipelines, and it’s quickly becoming the ubiquitous way to store data. Chapter 4 includes extensive coverage of Parquet, which includes how Parquet supports sophisticated object models such as Avro and how various Hadoop tools can use Parquet.”

Furthermore, “[h]ow data is being ingested into Hadoop has also evolved since the first edition,” Holmes points out, “and Kafka has emerged as the new data pipeline, which serves as the transport tier between your data producers and data consumers, where a consumer would be a system
such as Camus that can pull data from Kafka into HDFS. Chapter 5, which covers moving data into and out of Hadoop, now includes coverage of Kafka and Camus.”  [Reviewer’s note: Interesting software names here. Franz Kafka and Alfred Camus were writers deeply concerned about finding clarity and meaning in a world that seemed to offer none.]

Holmes adds that “[t]here are many new technologies that YARN now can support side by side in the same cluster, and some of the more exciting and promising technologies are covered in the new part 4, titled ‘Beyond MapReduce,’ where I cover some compelling new SQL technologies such as Impala and Spark SQL. The last chapter, also new for this edition, looks at how you can write your own YARN application, and it’s packed with information about important features to support your YARN application.”

Hadoop and MapReduce have gained reputations (well-earned) for being difficult to set up, use and master. In his new edition, Holmes describes his own early experiences: “The greatest challenge we faced when working with Hadoop, and specifically MapReduce, was relearning how to solve problems with it. MapReduce is its own flavor of parallel programming, and it’s quite different from the in-JVM programming that we were accustomed to. The first big hurdle was training our brains to think MapReduce, a topic which the book Hadoop in Action by Chuck Lam (Manning Publications, 2010) covers well.”

(These days, of course, there are both open source and commercial releases of Hadoop, as well as quickstart virtual machine versions that are good for learning Hadoop.)

Holmes continues: “After one is used to thinking in MapReduce, the next challenge is typically related to the logistics of working with Hadoop, such as how to move data in and out of HDFS and effective and efficient ways to work with data in Hadoop. These areas of Hadoop haven’t received much coverage, and that’s what attracted me to the potential of this book—the chance to go beyond the fundamental word-count Hadoop uses and covering some of the trickier and dirtier aspects of Hadoop.”

If you have difficulty explaining Hadoop to others (such as a manager or executive hesitant to let it be implemented), Holmes offers a succint summation in his updated book:

“Doug Cutting, the creator of Hadoop, likes to call Hadoop the kernel for big data, and I would tend to agree. With its distributed storage and compute capabilities, Hadoop is fundamentally an enabling technology for working with huge datasets. Hadoop provides a bridge between structured (RDBMS) and unstructured (log files, XML, text) data and allows these datasets to be easily joined together.”

One book cannot possibly cover everything you need to know about Hadoop, MapReduce, Parquet, Kafka, Camus, YARN and other technologies. And this book and its software examples assume that you have some experience with Java, XML and JSON. Yet Hadoop in Practice, Second Edition gives a very good and reasonably deep overview, spanning such major categories as background and fundamentals, data logistics, Big Data patterns, and moving beyond MapReduce.

Si Dunn



Hadoop is hot! Three new how-to books for riding the Big Data elephant – #programming #bookreview

In the world of Big Data, Hadoop has become the hard-charging elephant in the room.

Its big-name users now span the alphabet and include such notables as Amazon, eBay, Facebook, Google, the New York Times, and Yahoo. Not bad for software named after a child’s toy elephant.

Computer systems that run Hadoop can store, process, and analyze large amounts of data that have been gathered up in many different formats from many different sources.

According to the Apache Software Foundation’s Hadoop website: “The Apache Hadoop software library is a framework that allows for the distributed processing of large data sets across clusters of computers using simple programming models. It is designed to scale up from single servers to thousands of machines, each offering local computation and storage.”

The (well-trained) user defines the Big Data problem that Hadoop will tackle. Then the software handles all aspects of the job completion, including spreading out the problem in small pieces to many different computers, or nodes, in the distributed system for more efficient processing. Hadoop also handles individual node failures, and collects and combines the calculated results from each node.

But you don’t need a collection of hundreds or thousands of computers to run Hadoop. You can learn it, write programs, and do some testing and debugging on a single Linux machine, Windows PC or Mac. The Open Source software can be downloaded here. (Do some research first. You may have use web searches to find detailed installation instructions for your specific system.)

Hadoop is open-source software that is often described as “a Java-based framework for large-scale data processing.” It has a lengthy learning curve that includes getting familiar with Java, if you don’t already know it.

But if you are now ready and eager to take on Hadoop, Packt Publishing recently has unveiled three excellent how-to books that can help you begin and extend your mastery: Hadoop Beginner’s Guide, Hadoop MapReduce Cookbook, and Hadoop Real-World Solutions Cookbook.

Short reviews of each are presented below.

Hadoop Beginner’s Guide
Garry Turkington
(Packt Publishing – paperback, Kindle)

Garry Turkington’s new book is a detailed, well-structured introduction to Hadoop. It covers everything from the software’s three modes–local standalone mode, pseudo-distributed mode, and fully distributed mode–to running basic jobs, developing simple and advanced MapReduce programs, maintaining clusters of computers, and working with Hive, MySQL, and other tools.

“The developer focuses on expressing the transformation between source and result data sets, and the Hadoop framework manages all aspects of job execution, parallelization, and coordination,” the author writes.

He calls this capability “possibly the most important aspect of Hadoop. The platform takes responsibility for every aspect of executing the processing across the data. After the user defines the key criteria for the job, everything else becomes the responsibility of the system.”

The 374-page book is written well and provides numerous code samples and illustrations. But it  has one drawback for some beginners who want to install and  use Hadoop.  Turkington offers step-by-step instructions for how to perform a Linux installation, specifically Ubuntu. However, he refers Windows and Mac users to an Apache site where there is insufficient how-to information. Web searches become necessary to find more installation details.

Hadoop MapReduce Cookbook
Srinath Perera and Thilina Gunarathne
(Packt Publishing – paperback, Kindle)

MapReduce “jobs” are an essential part of  how Hadoop is able to crunch huge chunks of Big Data.  The Hadoop MapReduce Cookbook offers “recipes for analyzing large and complex data sets with Hadoop MapReduce.”

MapReduce is a well-known programming model for processing large sets of data. Typically, MapReduce is used within clusters of computers that are configured to perform distributed computing.

In the “Map” portion of the process, a problem is split into many subtasks that are then assigned by a master computer to individual computers known as nodes. (Nodes also can have sub-nodes). During the “Reduce” part of the task, the master computer gathers up the processed data from the nodes, combines it and outputs a response to the problem that was posed to be solved. (MapReduce libraries are now available for many different computer languages, including Hadoop.)

“Hadoop is the most widely known and widely used implementation of the MapReduce paradigm,” the two authors note.

Their 284-page book initially shows how to run Hadoop in local mode, which “does not start any servers but does all the work within the same JVM [Java Virtual Machine]” on a standalone computer. Then, as you gain more experience with MapReduce and the Hadoop Distributed File System (HDFS), they guide you into using Hadoop in more complex, distributed-computing environments.

Echoing the Hadoop Beginner’s Guide, the authors explain how to install Hadoop on Linux machines only.

Hadoop Real-World Solutions Cookbook
Jonathan R. Owens, Jon Lentz and Brian Femiano
(Packt Publishing – paperback, Kindle)

The Hadoop Real-World Solutions Cookbook assumes you already have some experience with Hadoop. So it jumps straight into helping “developers become more comfortable with, and proficient at solving problems in, the Hadoop space.”

Its goal is to “teach readers how to build solutions using tools such as Apache Hive, Pig, MapReduce, Mahout, Giraph, HDFS, Accumulo, Redis, and Ganglia.”

The 299-page book is packed with code examples and short explanations that help solve specific types of problems. A few randomly selected problem headings:

  • “Using Apache Pig to filter bot traffic from web server logs.”
  • “Using the distributed cache in MapReduce.”
  • “Trim Outliers from the Audioscrobbler dataset using Pig and datafu.” 
  • “Designing a row key to store geographic events in Accumulo.”
  • “Enabling MapReduce jobs to skip bad records.”

The authors use a simple but effective strategy for presenting problems and solutions. First, the problem is clearly described. Then, under a “Getting Ready” heading, they spell out what you need to  solve the problem. That is followed by a “How to do it…” heading where each step is presented and supported by code examples. Then, paragraphs beneath a “How it works…” heading sum up and explain how the problem was solved. Finally, a “There’s more…” heading highlights more explanations and links to additional details.

If you are a Hadoop beginner, consider the first two books reviewed above. If you have some Hadoop experience, you likely can find some useful tips in book number three

Si Dunn

MapReduce Design Patterns – For solving Big Data problems – #bookreview #programming #hadoop

MapReduce Design Patterns
Donald Miner and Adam Shook
(O’Reilly –
paperback, Kindle)

“MapReduce is a computing paradigm for processing data that resides on hundreds of computers,” the authors point out. It has been “popularized recently by Google, Hadoop, and many others.”

The MapReduce paradigm is “extraordinarily powerful, but does not provide a general solution to what many are calling ‘big data,” they add, “so while it works particularly well on some problems, some are more challenging.” The authors’ focus in their new book is on using MapReduce design patterns as “templates or general guides to solving problems.”

Their new book definitely can help solve some time-crunch problems for new MapReduce developers. It brings together a variety of solutions that have emerged over time in a patchwork of blogs, books, and research papers and explains them in detail, with code samples, illustrations, and cautions about potential pitfalls.

You can’t simply cut and paste solutions from the chapters. But the two writers do “hope that you will find a pattern to get you at least 90% of the way for just about all of your challenges.”

Six of the book’s eight chapters focus on specific types of design patterns:

  • Summarization Patterns
  • Filtering Patterns
  • Data Organization Patterns
  • Join Patterns
  • Metapatterns
  • Input and Output Patterns

“The MapReduce world is in a state similar to the object-oriented world before 1994,” the authors point out. “Patterns today are scattered across blogs, websites such as StackOverflow, deep inside other books, and inside very advanced technology teams at organizations across the world.”

They add that “[t]he intent of this book is not to provide some groundbreaking new ways to solve problems with MapReduce….” but to offer, instead, a collection of “patterns that have been developed by veterans in the field so they can be shared with everyone else.”

The book’s code samples are all written in Hadoop, and the two writers deal with the question of “why should we use Java MapReduce in Hadoop at all when we have options like Pig and Hive,” which reduce the need for MapReduce patterns.

There is “conceptual value,” they state, “in understanding the lower level workings of a system like MapReduce.” Furthermore, “Using Pig or Hive without understanding MapReduce can lead to some dangerous situations.” And, Pig and Hive cannot yet “tackle all of the problems in the ways that Java MapReduce can. This will surely change over time….”

If you are new to MapReduce, this useful and informative book from Donald Miner and Adam Shook can be the next best thing to having MapReduce experts at your side.

MapReduce Design Patterns can save you time and effort, steer you away from dead ends, and help give you solid understandings of the powerful MapReduce paradigm.

Si Dunn

Big Data Book Blast: Hadoop, Hive…and Python??? – #programming #bookreview

Big Data is hothotHOT. And O’Reilly recently has added three new books of potential interest to Big Data workers, as well as those hoping to join their ranks.

Hadoop, Hive and–surprise!—Python are just a few of the hot tools you may encounter in the rapidly expanding sea of data now being gathered, explored, stored, and manipulated by companies, organizations, institutions, governments, and individuals around the planet. Here are the books:

Hadoop Operations
Eric Sammer
(O’Reilly, paperbackKindle)

“Companies are storing more data from more sources in more formats than ever before,” writes Eric Sammer, a Hadoop expert who is principal solution architect at Cloudera. But gathering and stockpiling data is only “one half of the equation,” he adds. “Processing that data to produce information is fundamental to the daily operations of every modern business.”

Enter Apache Hadoop, a “pragmatic, cost-effective, scalable infrastructure” that increasingly is being used to develop Big Data applications for storing and processing information.

“Made up of a distributed filesystem called the Hadoop Distributed Filesystem (HDFS) and a computation layer that implements a processing paradigm called MapReduce, Hadoop is an open source, batch data processing system for enormous amounts of data. We live in a flawed world, and Hadoop is designed to survive in it by not only tolerating hardware and software failures, but also treating them as first-class conditions that happen regularly.”

Sammer adds: “Hadoop uses a cluster of plain old commodity servers with no specialized hardware or network infrastructure to form a single, logical, storage and compute platform, or cluster, that can be shared by multiple individuals or groups. Computation in Hadoop MapReduce is performed in parallel, automatically, with a simple abstraction for developers that obviates complex synchronization and network programming. Unlike many other distributed data processing systems, Hadoop runs the user-provided processing logic on the machine where the data lives rather than dragging the data across the network; a huge win for performance.”

Sammer’s new, 282-page book is well written and focuses on running Hadoop in production, including planning its use, installing it, configuring the system and providing ongoing maintenance. He also shows “what works, as demonstrated in crucial deployments.”

If you’re new to Hadoop or still getting a handle on it, you need Hadoop Operations. And even if you’re now an “old” hand at Hadoop, you likely can learn new things from this book. “It’s an extremely exciting time to get into Apache Hadoop,” Sammer states.

Programming Hive
Eric Capriolo, Dean Wampler, and Jason Rutherglen
(O’Reilly, paperback Kindle)

“Hive,” the three authors point out, “provides an SQL dialect, called Hive Query Language (abbreviated HiveQL or just HQL), for querying data stored in a Hadoop cluster.”

They add: “Hive is most suited for data warehouse applications, where relatively static data is analyzed, fast response times are not required, and when data is not changing rapidly.”

Their well-structured and well-written book shows how to install and test Hadoop and Hive on a personal workstation – “a convenient way to learn and experiment with Hadoop.” Then it shows “how to configure Hive for use on Hadoop clusters.”

They also provide a brief overview of Hadoop and MapReduce before diving into Hive’s command-line interface (CLI) and introductory aspects such as how to embed lines of comments in Hive v0.80 and later.

From there, the book flows smoothly into HiveQL and how to use its SQL dialect to query, summarize, and analyze large datasets that Hadoop has stored in its distributed filesystem.

User documentation for Hive and Hadoop has been sparse, so Programming Hive definitely fills a solid need. Significantly, the final chapter presents several “Case Study Examples from the User Trenches” where real companies explain how they have used Hive to solve some very challenging problems involving Big Data.

Python for Data Analysis
Wes McKinney
(O’Reilly, paperbackKindle)

No, Python is not the first language many people think of when picturing large data analysis projects. For one thing, it’s an interpreted language, so Python code runs a lot slower than code written in compiled programming languages such as C++ or Java.

Also, the author concedes, “Python is not an ideal language for highly concurrent, multithreaded applications, particularly applications with many CPU-bound threads.” The software’s global interpreter lock (GIL) “prevents the interpreter from executing more than one Python bytecode instruction at a time.”

Thus, Python will not soon be challenging Hadoop to a Big Data petabyte speed duel.

On the other hand, Python is reasonably easy to learn, and it has strong and widespread support within the scientific and academic communities, where a lot of data must get crunched at a reasonable clip, if not at blinding speed.

And Wes McKinney is the main author of pandas, Python’s increasingly popular open source library for data analysis. It (pandas) is “designed to make working with structured data fast, easy, and expressive.”

His book makes a good case for using Python in at least some Big Data situations. “In recent years,” he states, “Python’s improved library support (primarily pandas) has made it a strong alternative for data manipulation tasks. Combined with Python’s strength in general purpose programming, it is an excellent choice as a single language for building data-centric applications.”

Much of this well-written, well-illustrated book “focuses on high-performance array-based computing tools for working with large data sets.” It uses a case-study-examples approach to demonstrate how to tackle a wide range of data analysis problems, using Python libraries that include pandas, NumPy, matplotlib, and IPython, “the component in the standard scientific Python toolset that ties everything together.”

By the way, if you have never programmed in Python, check out the end of McKinney’s book. An appendix titled “Python Language Essentials” gives a good overview of the language, with a specific bias toward “processing and manipulating structured and unstructured data.”

If you do scientific, academic, or business computing and need to crunch and visualize a lot of data, definitely check out Python for Data Analysis.

You may be pleasantly surprised at how well and how easily Python and its data-analysis libraries can do the job.

Si Dunn