You can use CGroups to isolate CPU-heavy processes in a Hadoop cluster. If you are using CPU scheduling, you should also use CGroups to constrain and manage CPU processes. If you are not using CPU scheduling, do not enable CGroups.
When you enable CPU scheduling, queues are still used to allocate cluster resources, but both CPU and memory are taken into consideration using a scheduler that utilizes Dominant Resource Fairness (DRF). In the DRF model, resource allocation takes into account the dominant resource required by a process. CPU-heavy processes (such as Storm-on-YARN) receive more CPU and less memory. Memory-heavy processes (such as MapReduce) receive more memory and less CPU. The DRF scheduler is designed to fairly distribute memory and CPU resources among different types of processes in a mixed- workload cluster.
CGroups compliments CPU scheduling by providing CPU resource isolation. It enables you to set limits on the amount of CPU resources granted to individual YARN containers, and also lets you set a limit on the total amount of CPU resources used by YARN processes.
CGroups represents one aspect of YARN resource management capabilities that includes CPU scheduling, node labels, archival storage, and memory as storage. If CPU scheduling is used, CGroups should be used along with it to constrain and manage CPU processes.
