Book Image

Hadoop Operations and Cluster Management Cookbook

By : Shumin Guo
Book Image

Hadoop Operations and Cluster Management Cookbook

By: Shumin Guo

Overview of this book

<p>We are facing an avalanche of data. The unstructured data we gather can contain many insights that could hold the key to business success or failure. Harnessing the ability to analyze and process this data with Hadoop is one of the most highly sought after skills in today's job market. Hadoop, by combining the computing and storage powers of a large number of commodity machines, solves this problem in an elegant way!</p> <p>Hadoop Operations and Cluster Management Cookbook is a practical and hands-on guide for designing and managing a Hadoop cluster. It will help you understand how Hadoop works and guide you through cluster management tasks.</p> <p>This book explains real-world, big data problems and the features of Hadoop that enables it to handle such problems. It breaks down the mystery of a Hadoop cluster and will guide you through a number of clear, practical recipes that will help you to manage a Hadoop cluster.</p> <p>We will start by installing and configuring a Hadoop cluster, while explaining hardware selection and networking considerations. We will also cover the topic of securing a Hadoop cluster with Kerberos, configuring cluster high availability and monitoring a cluster. And if you want to know how to build a Hadoop cluster on the Amazon EC2 cloud, then this is a book for you.</p>
Table of Contents (15 chapters)
Hadoop Operations and Cluster Management Cookbook
Credits
About the Author
About the Reviewers
www.PacktPub.com
Preface
Index

Balancing data blocks for a Hadoop cluster


HDFS stores data blocks on DataNode machines. When Hadoop processes jobs, data is generated and deleted. Over time, some DataNodes can host much more data blocks than others. This unbalanced distribution of data on the cluster is called data skew .

Data skew is a big problem for a Hadoop cluster. We know that when the JobTracker assigns tasks to TaskTrackers, it follows the general rule of being data local , which means the map tasks will be assigned to those hosts where data blocks reside in. If the data block storage distribution is skewed, or in other words, the data blocks locate only on a small percentage of DataNodes, only those nodes with data blocks can follow the data local rule. Also, if JobTracker assigns tasks to other nodes that do not have data hosted locally, the data needs to be transferred from remote machines to the TaskTracker machine. The data transfer will cost a large amount of network bandwidth, downgrading the overall performance...