Using adaptive QoS policy groups
You can use an adaptive QoS policy group to automatically scale a throughput ceiling or floor to volume size, maintaining the ratio of IOPS to TBs|GBs as the size of the volume changes. That is a significant advantage when you are managing hundreds or thousands of workloads in a large deployment.
Before you begin
You must be running ONTAP 9.5. Adaptive QoS policy groups are available starting with ONTAP 9.5.
You must be a cluster administrator to create a policy group.
About this task
A storage object can be a member of an adaptive policy group or a non-adaptive policy group, but not both. The SVM of the storage object and the policy must be the same. The storage object must be online.
Adaptive QoS policy groups are always non-shared: the defined throughput ceiling or floor applies to each member workload individually.
The ratio of throughput limits to storage object size is determined by the interaction of the following fields:
- expected-iops is the minimum expected IOPS per allocated TB|GB.Noteexpected-iops is guaranteed on
AFA platforms only. expected-iops is guaranteed for FabricPool only if the tiering policy is set to none
and no blocks are in the cloud.expected-iops is guaranteed for volumes that are not in a SnapMirror Synchronous relationship. - peak-iops is the maximum possible IOPS per allocated or used TB|GB.
- expected-iops-allocation specifies whether allocated space (the default) or used space is used for expected-iops.Noteexpected-iops-allocation is available in ONTAP 9.5 and later.
- peak-iops-allocation specifies whether allocated space or used space (the default) is used for peak-iops.
- absolute-min-iops is the absolute minimum number of IOPS. You can use this field with very small storage objects. It overrides both peak-iops and/or expected-iops when absolute-min-iops is greater than the calculated expected-iops.
For example, if you set expected-iops to 1,000 IOPS/TB, and the volume size is less than 1 GB, the calculated expected-iops will be a fractional IOP. The calculated peak-iops will be an even smaller fraction. You can avoid this by setting absolute-min-iops to a realistic value.
- block-size specifies the application I/O block size. The default is 32K. Valid values are 8K, 16K, 32K, 64K, ANY. ANY means that the block size is not enforced.
Three default adaptive QoS policy groups are available, as shown in the following table. You can apply these policy groups directly to a volume.
Default policy group | Expected IOPS/TB | Peak IOPS/TB | Absolute Min IOPS |
---|---|---|---|
extreme | 6,144 | 12,288 | 1000 |
performance | 2,048 | 4,096 | 500 |
value | 128 | 512 | 75 |