Uber overcommits CPU resources to improve the efficiency of its computing infrastructure. To execute this strategy, Uber developed the Container Quality of Service (cQoS) system. cQoS collects per-container metrics, which improve scheduling decisions (including by identifying hot hosts) and detection of performance issues. The cQoS system includes a daemon that collects system metrics for the host and running containers and an advisor that ingests data from the daemons to find and analyze hot hosts. This system enabled Uber to reduce variance in CPU utilization across hosts in a cluster (through load-aware placement and by eliminating local hot spots) and address other performance bottlenecks.