We're assuming that most readers will have a little familiarity with Hadoop, or at the very least, with big data-processing systems. Consequently, we won't give a detailed background as to why Hadoop is successful or the types of problem it helps to solve in this book. However, particularly because of some aspects of Hadoop 2 and the other products we will use in later chapters, it is useful to give a sketch of how we see Hadoop fitting into the technology landscape and which are the particular problem areas where we believe it gives the most benefit.

In ancient times, before the term "big data" came into the picture (which equates to maybe a decade ago), there were few options to process datasets of sizes in terabytes and beyond. Some commercial databases could, with very specific and expensive hardware setups, be scaled to this level, but the expertise and capital expenditure required made it an option for only the largest organizations. Alternatively, one could build a custom system aimed at the specific problem at hand. This suffered from some of the same problems (expertise and cost) and added the risk inherent in any cutting-edge system. On the other hand, if a system was successfully constructed, it was likely a very good fit to the need.

Few small- to mid-size companies even worried about this space, not only because the solutions were out of their reach, but they generally also didn't have anything close to the data volumes that required such solutions. As the ability to generate very large datasets became more common, so did the need to process that data.

Even though large data became more democratized and was no longer the domain of the privileged few, major architectural changes were required if the data-processing systems could be made affordable to smaller companies. The first big change was to reduce the required upfront capital expenditure on the system; that means no high-end hardware or expensive software licenses. Previously, high-end hardware would have been utilized most commonly in a relatively small number of very large servers and storage systems, each of which had multiple approaches to avoid hardware failures. Though very impressive, such systems are hugely expensive, and moving to a larger number of lower-end servers would be the quickest way to dramatically reduce the hardware cost of a new system. Moving more toward commodity hardware instead of the traditional enterprise-grade equipment would also mean a reduction in capabilities in the area of resilience and fault tolerance. Those responsibilities would need to be taken up by the software layer. Smarter software, dumber hardware.

Google started the change that would eventually be known as Hadoop, when in 2003, and in 2004, they released two academic papers describing the Google File System (GFS) (http://research.google.com/archive/gfs.html) and MapReduce (http://research.google.com/archive/mapreduce.html). The two together provided a platform for very large-scale data processing in a highly efficient manner. Google had taken the build-it-yourself approach, but instead of constructing something aimed at one specific problem or dataset, they instead created a platform on which multiple processing applications could be implemented. In particular, they utilized large numbers of commodity servers and built GFS and MapReduce in a way that assumed hardware failures would be commonplace and were simply something that the software needed to deal with.

At the same time, Doug Cutting was working on the Nutch open source web crawler. He was working on elements within the system that resonated strongly once the Google GFS and MapReduce papers were published. Doug started work on open source implementations of these Google ideas, and Hadoop was soon born, firstly, as a subproject of Lucene, and then as its own top-level project within the Apache Software Foundation.

Yahoo! hired Doug Cutting in 2006 and quickly became one of the most prominent supporters of the Hadoop project. In addition to often publicizing some of the largest Hadoop deployments in the world, Yahoo! allowed Doug and other engineers to contribute to Hadoop while employed by the company, not to mention contributing back some of its own internally developed Hadoop improvements and extensions.