Storage I/O Control Overview
- SIOC is a method to ensure that the VM's on datastore get a fair share of I/O resources during the period of high contention or latency.
- Important Virtual Machines get preferences when compared to less important Virtual Machines for I/O resource allocation.
- Can be used both with and without Storage DRS as it includes two different thresholds i.e Standalone without DRS and with vSphere Storage DRS.
- SIOC is aware about the average latency, active VM's and also the priorities associated with them.
- Quite Similar to the concept used by DRS for CPU/Memory the higher the shares value higher the resources it get.
- vCenter is only used for enabling and disabling the feature and for changing the thresholds.
- SIOC is supported on FC,ISCSI, and NFS not supported with Raw Device Mapping (RDM)
- vSphere Enterprise Plus License.
- Datastore which are SIOC enabled must be managed by single vCenter Server.
How It Works?
- To configure the vSphere SIOC enable it for the particular datastore as it is disabled by default.
- Set the number of storage I/O shares for each VM's and limit for IOPS by default the limit is unlimited.
- Once Enabled ESXi starts monitoring the latency when communicating with the datastore and if it exceeds the threshold value means it's congested.
- SIOC uses Workload Injector Based Model and only runs when the datastore is in an ideal state.
- It automatically determines the latency, moreover this can also be overridden by specifying specific latency value.
- Automatic detection works well when you have multiple disk arrays and many datastore are configured for SIOC.
For More information refer vSphere Documentation Center