Goal

VMware provides esxtop, a tool customers use to analyze the storage performance of a VM. The CNV team would like OCP-v customers to have a similar tool, namely one that reports IOPS, IO latency, and IO queue depth. Currently, the only one missing is IO queue depth, which is the purpose of this API request. 

I have done some benchmarking of Qemu's queue depth metric and found that it is reasonably accurate compared to iostat's reported queue depth and also only uses marginal cpu overhead. Therefore, if Libvirt had an API to report Qemu's queue depth, then Kubevirt could track the queue depth with Prometheus, and the OCP-v web console could graph it.  

Customers would then be able to diagnose storage performance issues the same way they would be able to on VMware.

Related tickets:  RHEL-104408 CNV-60441