Book Image

Practical Real-time Data Processing and Analytics

Book Image

Practical Real-time Data Processing and Analytics

Overview of this book

With the rise of Big Data, there is an increasing need to process large amounts of data continuously, with a shorter turnaround time. Real-time data processing involves continuous input, processing and output of data, with the condition that the time required for processing is as short as possible. This book covers the majority of the existing and evolving open source technology stack for real-time processing and analytics. You will get to know about all the real-time solution aspects, from the source to the presentation to persistence. Through this practical book, you’ll be equipped with a clear understanding of how to solve challenges on your own. We’ll cover topics such as how to set up components, basic executions, integrations, advanced use cases, alerts, and monitoring. You’ll be exposed to the popular tools used in real-time processing today such as Apache Spark, Apache Flink, and Storm. Finally, you will put your knowledge to practical use by implementing all of the techniques in the form of a practical, real-world use case. By the end of this book, you will have a solid understanding of all the aspects of real-time data processing and analytics, and will know how to deploy the solutions in production environments in the best possible manner.
Table of Contents (20 chapters)
Title Page
Credits
About the Authors
About the Reviewers
www.PacktPub.com
Customer Feedback
Preface

State retention and the need for Trident


Trident is a distributed real-time analytics framework. Trident maintains its state either internally for example, in-memory, or externally for example, Hazelcast, in a fault-tolerant way. It is similar to processing an event exactly once. Trident fits for micro batch processing use cases such as aggregation, filtration, and so on.

Let's take an example that explains how to achieve exactly-once semantics. Suppose that you're doing a count of how many people visited your blog and also storing the running count in a database. Now suppose you store a single value representing the count in the database, and every time you process a new tuple you increment the count.

Now, if failures happen, tuples will be replayed by Storm topology. Here the problem is whether or not the tuple has been processed and the count has already been updated in the database—if so, then you should not update it again or if the tuple did not process successfully then you have to...