SIOC provides quality of service capabilities for storage I/O in the form of I/O shares and limits that are enforced across all the VMs accessing a datastore, regardless of which host they are running on. Using SIOC, vSphere administrators can ensure that the most important VMs get adequate I/O resources, even in times of congestion.

In vSphere 5.1, SIOC can automatically determine the optimal latency threshold using injector-based models to determine the latency setting. In vSphere 5.1 and above, this injector determines and sets the latency threshold when 90 percent of the throughput is reached.

When SDRS is enabled, SIOC is set to stats only mode by default. Stats only mode collects and stores statistics but does not perform throttling on the storage device. Storage DRS can use the stored statistics immediately after the initial configuration or when new data stores are added.

When you enable SIOC on a datastore, ESXi begins to monitor the device latency that hosts observe when communicating with that data store. When device latency exceeds a threshold, the data store is considered to be congested, and each VM that accesses this data store is allocated I/O resources in proportion to their shares.

When you allocate storage I/O resources, you can limit the IOPS that are allowed for a VM. By default, the number of IOPS allowed for a VM is unlimited. If the limit that you want to set for a VM is in terms of megabytes per second instead of IOPS, you can convert megabytes per second into IOPS based on the typical I/O size of that VM. For example, a backup application has a typical I/O size of 64 KB. To restrict a backup application to 10 MB per second, set a limit of 160 IOPS (10 MB per second per 64 KB I/O size, which is equal to 160 I/OS per second). However, this setting needs an eye on it as an application with a 4 KB I/O size would be very slow if limited to 160 IOPS. An IOPS limit is applied to the sum of the limits of all the virtual disks attached to a VM.