How the maximum throughput limit works
You can specify one service-level objective for a Storage QoS policy group: a maximum throughput limit. A maximum throughput limit, which you define in terms of IOPS, MBps, or both, specifies the throughput that the workloads in the policy group cannot collectively exceed.
When you specify a maximum throughput for a policy group, Storage QoS controls client operations to ensure that the combined throughput for all workloads in the policy group does not exceed the specified maximum throughput.
For example, assume that you create the policy group “untested_apps” and specify a maximum throughput of 300 MBps. You assign three volumes to the policy group. The combined throughput to those three volumes cannot exceed 300 MBps.
The combined throughput to the workloads in a policy group might exceed the specified limit by up to 10 percent. A deviation might occur if you have a workload that experiences rapid changes in throughput (sometimes called a bursty workload).
Note the following about specifying a maximum throughput:
You must not set the limit too low because you might underutilize the cluster.
You must consider the minimum amount of throughput that you want to reserve for workloads that do not have limits.
For example, you can ensure that your critical workloads get the throughput that they need by limiting noncritical workloads.
You might want to provide room for growth.
For example, if you see an average utilization of 500 IOPS, you might specify a limit of 1,000 IOPS.