Spark Core

Abstract

Spark is the most active open source project in the big data world. It has become hotter than Hadoop. It is considered the successor to Hadoop MapReduce, which we discussed in Chapter 1. Spark adoption is growing rapidly. Many organizations are replacing MapReduce with Spark. Conceptually, Spark looks similar to Hadoop MapReduce; both are designed for processing big data. They both enable cost-effective data processing at scale using commodity hardware. However, Spark offers many advantages over Hadoop MapReduce. These are discussed in detail later in this chapter. This chapter covers Spark core, which forms the foundation of the Spark ecosystem. It starts with an overview of Spark core, followed by the high-level architecture and runtime view of an application running on Spark. The chapter also discusses Spark core's programming interface. Overview Spark is an in-memory cluster computing framework for processing and analyzing large amounts of data. It provides a simple programming interface, which enables an application developer to easily use the CPU, memory, and storage resources across a cluster of servers for processing large datasets. Key Features The key features of Spark include the following: • Easy to use • Fast • General-purpose • Scalable • Fault tolerant Easy to Use Spark provides a simpler programming model than that provided by MapReduce. Developing a distributed data processing application with Spark is a lot easier than developing the same application with MapReduce. Spark offers a rich application programming interface (API) for developing big data applications; it comes with 80-plus data processing operators. Thus, Spark provides a more expressive API than that offered by Hadoop MapReduce, which provides just two operators: map and reduce. Hadoop MapReduce requires every problem to be broken down into a sequence of map and reduce jobs. It is hard to express non-trivial