Apache Spark Cluster (Python stack with Jupyter notebook and PySpark)

Apache Spark is a fast and general-purpose cluster computing system. It provides high-level APIs in Java, Scala, Python and R, and an optimized engine that supports general execution graphs. It also supports a rich set of higher-level tools including Spark SQL for SQL and structured data processing, MLlib for machine learning, GraphX for graph processing, and Spark Streaming. For more information visit the official Apache Spark page.

Apache Spark cluster together with HDFS (Hadoop Distributed File System) represents one of the most important tool for Big Data and machine learning applications, enabling the parallel processing of large data sets on many virtual machines, which are running Spark workers. On the other hand, setting up a Spark cluster with HDFS on clouds is not straightforward, requiring deep knowledge of both cloud and Apache Spark architecture. To save this hard work for scientists we have created and made public the required infrastructure descriptors by which Occopus can automatically deploy Spark clusters with the number of workers specified by the user. Spark also provides a special library called “Spark MLlib” for supporting machine learning applications. Similarly, to the R-oriented Spark environment, we have developed the infrastructure descriptors for the creation of a machine learning environment in the cloud. Here, the programming language is Python and the user programming environment is Jupyter. The complete machine learning environment consists of the following components: Jupyter, Python, Spark and HDFS. Deploying this machine learning environment is also automatically done by Occopus and the number of Spark workers can be defined by the user.

This tutorial sets up a complete Apache Spark infrastructure integrated with HDFS, Python and Jupyter Notebook. It contains a Spark Master node and Spark Worker nodes, which can be scaled up or down.