I'm in desperate need of scalable Apache Spark cluster available through API that would make it easy to submit jobs that could process arbitrary size datasets but would let me abstract away the scaling part of the problem. I don't understand how there's nothing like that already considering popularity of Spark.