In Chapter 4, Predicting User Behavior with Tree-Based Methods, and other chapters, we had to launch Elastic MapReduce (EMR) clusters and SageMaker instances (servers) for learning and model serving. In this section, we discuss the characteristics of the different instance types. In this chapter, you can find all the supported instance types that AWS provides at https://aws.amazon.com/ec2/instance-types/.
Depending on the task at hand, we should use different instance types. For example, we may require an instance type with graphical processing units (GPUs), rather than CPUs, for deep learning. When launching a large iterative extract, transform, and load (ETL) job (that is, a data transformation job) on Apache Spark, we might need large amounts of memory. To make it easier for the users, AWS has classified the instances into families that are catered...