QoS provides upper-limit control for volumes and QoS volume groups. When a host I/O request* is issued to a volume or a QoS group for which an upper-limit value is set, the storage system checks the I/O rate or transfer rate of all volumes or QoS groups for the last 1 second. If the throughput reaches the upper-limit value, the storage system suppresses host I/O processing for the volume and, if the volume belongs to a QoS group, for all volumes in the QoS group to which the volume belongs, while it continues to accept I/O requests from services. When the throughput falls below the upper-limit value, the storage system resumes I/O processing for the volume and, if the volume belongs to a QoS group, for all volumes in the QoS group.
* The QoS function controls the upper limit only for I/Os between the hosts and the storage system. When QoS-controlled volumes are used as pair volumes for remote copy operations, the QoS upper limit does not apply to remote I/Os between the primary and secondary storage systems.
QoS upper-limit control can be applied to the I/O rate, the transfer rate, or both. If upper-limit control is applied to both the I/O rate and the transfer rate, I/O processing is suppressed when either the I/O rate or the transfer rate reaches the upper-limit value. You can set the following values for upper-limit control of volumes and QoS groups:
- I/O rate (IOPS): 100 to 2,147,483,647
- Transfer rate (MB/s): 1 to 2,097,151
You can use upper-limit control in combination with lower-limit control and priority control. For more information, see Combining QoS upper-limit, lower-limit, and priority controls for a volume.
- The upper limit for a QoS group is shared by all volumes in the QoS group.
- I/O operations are performed once per second for all volumes that receive host I/O requests even after the throughput has reached the upper limit. As a result, I/O operations are temporarily performed exceeding the upper limit, but the performance degradation of the host due to the I/O suppression can be suppressed.
- A volume that has its own individual upper limit can also be included in a QoS group. For more information, see Combinations of the QoS settings for a volume and a QoS group.
- When a volume with an upper limit set is added to a QoS group, or when an upper limit for a volume included in a QoS group is changed, if either the upper limit to be set for the volume or that to be set for the QoS group is reached, I/O operations are suppressed. If the I/O performance is different from the expected value, verify both the upper limit for the volume and the upper limit for the QoS group.
The following table describes the differences between using a QoS group and not using a QoS group for upper-limit control.
|
Condition |
QoS upper limit settings |
|---|---|
|
If QoS group is not used and the same upper limit (for example, 1,000 IOPS) is set for five volumes |
The upper limit for each volume remains 1,000 IOPS and it is not changed. |
|
If five volumes with no upper limits set are grouped into a QoS group, and 5,000 IOPS is set as an upper limit for the QoS group |
The five volumes share 5,000 IOPS.
|
|
If a volume with the upper limit of 2,000 IOPS set and four volumes with no upper limits set are grouped into a QoS group, and 6,000 IOPS is set as an upper limit for the QoS group |
The five volumes share 6,000 IOPS.
|