Both pool types use the same node size table; a Spark instance always has one head node (running the driver, Livy, YARN Resource Manager) and one or more worker nodes running the Spark Executor ...