Comment by matteuan

We use Lambda with S3 as intermediate storage between different steps, sort of multiple-stages map-only MapReduce. And we still use Hadoop on premise :)