To process continuous streams of data from sources HDFS directories, TCP sockets, Kafka, Flume, Twitter e.t.c spark has two methodologies.
Spark streaming in general works on something called as a micro batch. The stream pipeline is registered with some operations and the Spark polls the source after every batch duration (defined in the application) .
Spark streaming uses DStream api and spark structured streaming uses dataset/dataframe — short answer :-) .