Database Reference
In-Depth Information
Figure 6.23
VMware vSphere storage queues.
Caution
If you have an under-configured storage array and insufficient individual spindles
or disks to service the aggregate IO requirements, then increasing the queue depth
will not improve performance. On an under-configured array, increasing queue
depth will just result in the queues becoming full and increased IO latency or
service times. Virtualization doesn't get around the laws of physics; you may
need more disks. Our goal is to ensure the path from the guest to the underlying
storage is not the bottleneck in software, so that you can get the most out of your
physical storage investments and get the highest performance possible.
Note
If you are presenting RDMs to a VM, such as with SQL FCI, and using the
LSI Logic SAS adapter, there is little benefit in increasing the queue depth
to 64. Windows will only be able to issue 32 outstanding IOs before it
starts queuing, and you'll never be able to make use of the additional queue
depth. If you will be using a large number of RDMs on your hosts, see
VMware KB 1016106.
