The Spark driver plans and coordinates the set of tasks required to run a Spark application. The tasks themselves run in executors, which are hosted on worker nodes.