As seen previously, Spark Streaming uses a concept of DStreams, which are essentially micro-batches of data created as RDDs. We also saw types of transformations that are possible on DStreams. The transformations on DStreams can be grouped into two types: Stateless transformations and Stateful transformations.

In Stateless transformations, the processing of each micro-batch of data does not depend on the previous batches of data. Thus, this is a stateless transformation, with each batch doing its own processing independently of anything that occurred prior to this batch.

In Stateful transformations, the processing of each micro-batch of data depends on the previous batches of data either fully or partially. Thus, this is a stateful transformation, with each batch considering what happened prior to this batch and then using the information while computing the data in this batch.