Mark Grover

This is a book for software / data engineers who've been using Hadoop and related technologies for a while in practical projects, as well as for software architects looking for high level overview of how many of Big Data technology stack components relate to each other, and justifications to use which of them in different use cases.

The book is very well and clearly organized, and proceeds very logically in terms of Hadoop storage options, how to put / ingest data into a Hadoop environment, how to decide and use processing engines for Hadoop such as MapReduce, Spark, Hive, etc., how to utilize those engines to do important and critical tasks such as record deduplication, windowing analysis, and time series modification. The exposition of these fundamental building blocks are followed by graph processing on Hadoop, where both Giraph and Spark GraphX are described and contrasted. And then the topic of orchestration of Hadoop workflows are described to an extent, mainly showing how to configure and use Oozie. Part I finishes by describing Near-Realtime processing in Hadoop, and shows how Storm, Trident and Spark Streaming can be used for satisfying different requirements.

The second part of the book is dedicated to real-world use cases such as Clickstream Analytics, Fraud Detection, and Data Warehousing. The authors provide a good and broad overview for each case, clearly showing where and how Hadoop software stack helps, together with architectural recommendations, but I think the the final use case, Data Warehouse chapter is the most interesting one because it makes use of a very popular, publicly available movie data set known as MovieLens. Thanks to this, it is very easy to follow this chapter by using the same data and apply the designs and programming steps, creating your own customizations and investigating different scenarios and technical challenges you can come up with.

As a conclusion, I can recommend this book to big data architects and software engineers who are not total novices when it comes to Hadoop. The book is of course a bit date, in the very fast moving world of big data, 2015 sounds already distant past, but thanks to the extensive industrial and practical experience of authors, the way they explain their thinking and justifications for very different scenarios shed light on current and upcoming challenges for many big data engineers.… (plus d'informations)