They are meant to provide easier concurrency mechanism. A way to coordinate concurrent routines. They are not meant to build a data pipeline for processing. I agree that Heka's use of channel was not optimal.