YARN manages cluster resources by using queues. Resources like CPU and memory are partitioned into queues, to which applications can then be assigned to run in. Each YARN queue is allocated a certain amount of CPU and memory (fixed or elastic), which is shared among the applications assigned to that queue.