Priority Optimization policy settings

Policy priority – Priority Optimization policies can have a priority of high, normal, or low. When a storage system gets busy, it can start to throttle (reduce) the IOPS, bandwidth, or both, of the virtual volumes that are covered by the lower priority policies. This is done to meet the latency goals of the higher priority policies.

  • A priority setting of high should be used for critical applications. A setting of normal and low priority should be used for less critical applications.

  • The priority settings are ignored if a storage system does not have any policies that specify a latency goal, IOPs minimum performance goal, or a bandwidth minimum performance goal.

Latency goal – A policy can specify a latency time goal (ms) for the virtual volumes that are covered by the policy. The latency goal is also known as a response time goal.

  • If a latency goal is specified in a policy, other policies must specify minimum performance goals so that the storage system can determine how much it can throttle the IOPS and bandwidth of the virtual volumes that are covered by the lower priority policies.

Maximum performance limits – A policy can specify the maximum performance limit for the virtual volumes in the policy target. A storage system throttles (reduces) the performance to prevent the maximum limits from being exceeded. Maximum performance limits can be specified for IOPS (IO/s), bandwidth (KB/s), or both.

Minimum performance goals – A policy can specify the minimum performance goal for the virtual volumes in the policy target. A storage system will not throttle (reduce) performance below the minimum goals in order to meet the latency goal of a higher priority policy. Minimum performance goals can be specified for IOPS (IO/s), bandwidth (KB/s), or both.

  • If a policy specifies minimum performance goals for IOPS or bandwidth, then a maximum performance limit for IOPS or bandwidth must also be specified in the policy.

  • IOPS and bandwidth minimum performance goals are ignored if the storage system does not have any policies that specify a latency goal.