An ERP application VM on a densely utilized virtualization host slows down every afternoon. During the slowdown the administrator records these metrics from monitoring tools: CPU utilization 35 %, CPU ready 12 %, memory utilization 52 %, average disk latency 3 ms, and network throughput 120 Mbps on a 1 Gbps NIC. Based on these performance metrics, which resource is most likely the bottleneck that is affecting the VM's response time?
Insufficient physical memory causing paging because memory utilization is at 52 %
Network interface saturation because throughput is already 120 Mbps on a 1 Gbps NIC
Excessive vCPU scheduling contention indicated by the high CPU ready time
Storage latency; the 3 ms average disk response shows the disks cannot keep up
CPU ready reports how long a vCPU is "ready" to run but cannot get time on a physical processor. Values above roughly 10 % are widely accepted as symptomatic of CPU contention on an oversubscribed host. In this scenario the VM shows CPU ready at 12 %, well past the common 5 - 10 % comfort range, even though its CPU utilization is only 35 %. The other metrics do not indicate stress: 3 ms disk latency is low for modern storage, memory utilization at 52 % leaves plenty of free RAM, and 120 Mbps uses only about 12 % of a 1 Gbps link. Therefore vCPU scheduling contention on the host is the most probable cause of the afternoon slow-downs, while the remaining resources are unlikely bottlenecks.