Where is cached RDD stored (i.e. in a distributed way or on a single node)?

When we cache an RDD in Spark, is it then stored in a distributed way or on a single node? In which systems memory is it stored?