Unlike the Spark transformer, the stream transformer reads data directly from the enriched stream and does not use Spark or EMR. It's a plain JVM application, like Stream Enrich or S3 Loader.
Reading directly from stream means that the transformer can bypass the
s3DistCp staging / archiving step.
Another benefit is that it doesn't process a bounded data set and can emit transformed folders based only on its configured frequency. This means the pipeline loading frequency is limited only by the storage target.