What is the purpose of RDD transformations in Apache Spark?

The choice that states RDD transformations are used to create new RDDs from existing ones is accurate. RDD, or Resilient Distributed Dataset, is a fundamental data structure in Apache Spark that provides a resilient and distributed way to process large amounts of data.

Transformations in Spark, such as map, filter, and flatMap, apply to RDDs and produce new RDDs by performing operations on the data in the existing RDDs. These transformations are lazy, meaning they do not execute immediately. Instead, they define a lineage of operations that can be computed when an action is invoked. This behavior allows for optimization, as Spark can build an efficient execution plan based on the transformations that are defined.

When performing these transformations, the original RDD remains unchanged, which aligns with the functional programming paradigm and immutability principles. This means that RDD transformations preserve the history of operations, enabling Spark to more effectively manage and recover from failures.

The other options do not accurately reflect the purpose of transformations. Permanent changes to data would pertain to actions that modify the data in a persistent storage system. Immediate computations relate to actions, which trigger the execution of the data processing pipeline. Manipulating metadata doesn't align with the